Hi Charles,
I also looked at the IRF P-channel transconductance anomaly, but only after someone (probably Nelson) brought it to my attention. I did quite a bit of lab experimentation and also contacted people at IR to see if I could find out more about it. They were of no help. I never did figure out what was the cause of it, although I don't remember think it was thermal. Seems I remember thinking it was something akin to some trapped charge under the gate with a slow discharge path.
If the problem were thermal, it could lead to some form of memory distortion, but I think one could arrange experiments looking at time constants of activity to sort out whether it was thermal or not.
I think someone told me that the Vishays (used to be IR) surprizingly did not suffer the effect, even though one might have expected that they would have just ported over the same process when they bought IRs MOSFET business.
P-channel MOSFETs will never really match N-channel MOSFETs (nor much of any other semiconductor), so it is something we have to live with and just mitigate as best we can. In my book I show how to achieve better balance in the crossover region by using asymmetrical source resistors of relatively small values.
The nonlinear capacitance of the verticals is a PIA in some cases if you drive the output close to the rails, but BJT power transistors also have highly nonliear Ccb at low Vcb and of a similar order of magnitude.
The Toshiba verticals are great devices, but apparently no longer available.
As I mention in my book, I prefer verticals to laterals, and I think there are some good reasons for that. Nevertheless, one can also make a good amplifier out of laterals. The Vgs TC of the laterals is nice, of course, but that of the verticals is still far easir to manage than that of BJTs. ThermalTrak finally to the rescue! Laterals also are harder to blow up, but that is partly due to their lower gm, lower speed, high Rds and often built-in gate-source zeners.
Cheers,
Bob
Hi,
Unsurprisingly.
Laterals can be safely biased to run quite hot, at around 400mA per device the "wingspread diagram" flattens out nicely, in ways similar to a good Class AB PP Tube powerstage (not straightline, but not bad).
It is not possible to get these results without ton's of extra circuitry (or more feedback than is stable and feedback is kinda a bad thing in my religion anyway) and then success is uncertain, if using Verticals or BJT's (and IGBT's).
I actually kinda can now see why some people throught these where good in the 80's. In a sensible arrangement they do some things nothing else available does and the output stage largely determines overall distortion and speed (HF response) in any kind of Amp (tube or transistor).
On the other hand, I picked up a more than healthy bias against laterals in the 80's. Ton's of cheapish PA Amp's that came out at the time, with laterals, all based on the original Hitatchi App note Circuit.
Fitted with way too few pairs of output pairs (cost) for the power, thus driving up output impedance when running hot (or even when not) and killing LF power output, no drivers (reliability) and way too low VAS current to drive the existing output pairs input capacitance well causing slewing every time the drummer hit the crash cymbals hard, with the result that bass turned into a soggy undefined mush and the treble was awful, we called it Mosfet Mist (fog in english, animal excrement in german, very funny)...
I did not really understand the why at the time (hind sight is 40/40) and it foxed the heck out of me, it just sounded so total pants, useless...
As we had the Amps and paid for them as well, so we kept them on the mids where they sounded quite okay. We kept bipolars (mostly Peavey CS series) in the bass and treble, until I perfected my "Borg Cubes" (they do predate the Borg on Star Trek, but they where very powerful cubes, so they acquired the ticket later).
The Borg Cubes where also all bipolar, all NPN, lowish feedback circlotrons, transistors either industrial 300A/380V darlington switchers (outputs) or TV Vertical Deflection (single stage VAS & Drivers). Power was nearly 1KW RMS into 8Ohm, collapsing into lower loads to safe levels for Electro Voice Pro-Line Drivers (well, almost safe),
Originally they where surprisingly zero feedback (I did not know how to apply feedback in a fully balanced circuit at the time), but all devices had so much beta-drop that while sounding really great with a single or two pairs of 8 Ohm Full Range 15/3 Box(s), they lost it into 8pcs of dual 18" W-Bin Bass Horns, though not as bad as the mosfet amp's into a pair of 8 Ohm full range speakers in parallel.
So I added another stage and around 30dB of feedback(took me a few month), it "fixed" the muddy bass, but they now only sounded great on bass bin's and lower mid horns, the upper midrange and treble went from so-so midrange to not great treble...
So we bought some "Zeck BiMos" Amp's (mosfet driver, bipolar outs) that sounded great on the 110dB+/1W/1m Horns we used on the upper mids and treble.
Now remember, we are talking 1980's rock band large hall/open air Sound Reinforcement systems. We used to drive the drivers within a few inches of their life and then some.
LIKE VERY LOUD.
I realise now, the datasheets I had on all the parts where illustrations of Bob Pease's comments and I was too inexperienced to jury-rig some test setup to characterise the devices for my own applications, I was young, ignorant and way too randy back then to work it out, the Amp's worked (Kinda) and there where girls to chase and more...
So, I am also a (weary and tentative) convert to Hitachi/Renesas Laterals. Well I bought a few, I hope the amplifier mods I have in mind work out...
Then again, maybe a correctly biased lateral driver Sziklay would be a better choice? Only one way to find out, build it and listen...
Ciao T
Unsurprisingly.
Laterals can be safely biased to run quite hot, at around 400mA per device the "wingspread diagram" flattens out nicely, in ways similar to a good Class AB PP Tube powerstage (not straightline, but not bad).
It is not possible to get these results without ton's of extra circuitry (or more feedback than is stable and feedback is kinda a bad thing in my religion anyway) and then success is uncertain, if using Verticals or BJT's (and IGBT's).
I actually kinda can now see why some people throught these where good in the 80's. In a sensible arrangement they do some things nothing else available does and the output stage largely determines overall distortion and speed (HF response) in any kind of Amp (tube or transistor).
On the other hand, I picked up a more than healthy bias against laterals in the 80's. Ton's of cheapish PA Amp's that came out at the time, with laterals, all based on the original Hitatchi App note Circuit.
Fitted with way too few pairs of output pairs (cost) for the power, thus driving up output impedance when running hot (or even when not) and killing LF power output, no drivers (reliability) and way too low VAS current to drive the existing output pairs input capacitance well causing slewing every time the drummer hit the crash cymbals hard, with the result that bass turned into a soggy undefined mush and the treble was awful, we called it Mosfet Mist (fog in english, animal excrement in german, very funny)...
I did not really understand the why at the time (hind sight is 40/40) and it foxed the heck out of me, it just sounded so total pants, useless...
As we had the Amps and paid for them as well, so we kept them on the mids where they sounded quite okay. We kept bipolars (mostly Peavey CS series) in the bass and treble, until I perfected my "Borg Cubes" (they do predate the Borg on Star Trek, but they where very powerful cubes, so they acquired the ticket later).
The Borg Cubes where also all bipolar, all NPN, lowish feedback circlotrons, transistors either industrial 300A/380V darlington switchers (outputs) or TV Vertical Deflection (single stage VAS & Drivers). Power was nearly 1KW RMS into 8Ohm, collapsing into lower loads to safe levels for Electro Voice Pro-Line Drivers (well, almost safe),
Originally they where surprisingly zero feedback (I did not know how to apply feedback in a fully balanced circuit at the time), but all devices had so much beta-drop that while sounding really great with a single or two pairs of 8 Ohm Full Range 15/3 Box(s), they lost it into 8pcs of dual 18" W-Bin Bass Horns, though not as bad as the mosfet amp's into a pair of 8 Ohm full range speakers in parallel.
So I added another stage and around 30dB of feedback(took me a few month), it "fixed" the muddy bass, but they now only sounded great on bass bin's and lower mid horns, the upper midrange and treble went from so-so midrange to not great treble...
So we bought some "Zeck BiMos" Amp's (mosfet driver, bipolar outs) that sounded great on the 110dB+/1W/1m Horns we used on the upper mids and treble.
Now remember, we are talking 1980's rock band large hall/open air Sound Reinforcement systems. We used to drive the drivers within a few inches of their life and then some.
LIKE VERY LOUD.
I realise now, the datasheets I had on all the parts where illustrations of Bob Pease's comments and I was too inexperienced to jury-rig some test setup to characterise the devices for my own applications, I was young, ignorant and way too randy back then to work it out, the Amp's worked (Kinda) and there where girls to chase and more...
So, I am also a (weary and tentative) convert to Hitachi/Renesas Laterals. Well I bought a few, I hope the amplifier mods I have in mind work out...
Then again, maybe a correctly biased lateral driver Sziklay would be a better choice? Only one way to find out, build it and listen...
Ciao T
This is going way back, but my friend who is an apps engineer at Siemens (now Infineon) knew what the problem was. I vaguely recall something about "diffusion" and that the problem had to do with something in the process itself. At that time it was over my head (probably still would be!) so it didn't stick with me.
That is possible. I have moved on from vertical MOSFETs and have no reason to measure them (see below).
The only disadvantage of a lateral compared to a vertical is the lower transconductance.
If you are using them in your drivers, JOHN CURL, then this is not a problem at all. Switch over to the Semelab parts and they will make your amps sound far better. (The Renesas P-channel parts are pretty weird.)
If you are using them as output devices, then they will have too high an output impedance. Just parallel a bunch. For an equivalent transconductance, the lateral parts will have less than half the input capacitance (Cdg, as we are talking about a follower here) of a vertical part. And near clipping, the input capacitance will be less than one-tenth that of a vertical part.
So you see, the lateral part is far superior in all ways for audio. The only disadvantage is cost. If you have to parallel four to eight devices to replace one vertical device, then it is going to drive the price way up.
If, on the other hand, you already have a boatload of vertical devices, then just replace them with lateral devices. There is no audio application that needs more than two or three vertical devices in parallel, so if you have eight or ten or twelve, then the lateral parts will just drop in. Change the values of the gate resistors, and possibly the source resistors. The bias circuit will simplify radically. The sound quality will improve radically.
Last edited:
Well I seem to have just had my CD player fail. So it is either FM or fix it. Cleaned my desk of work quicker than usual so I can have the time!
While at it I may get the chance to play with some of my IR Mosfet inventory to see what they do.
Bob, please for my sake could you limit references to your book to once per post!
You are beginning to make me feel like I should buy extra copies!
While at it I may get the chance to play with some of my IR Mosfet inventory to see what they do.
Bob, please for my sake could you limit references to your book to once per post!
You are beginning to make me feel like I should buy extra copies!Gentlemen, not only IR made VMOS, try to look at this ones.. IXYS Product Portfolio | Power Devices
There is really good " menu " for your taste...
There is really good " menu " for your taste...
Hi Upupa,
It's just that IR made many parts at affordable prices. A strong motivation when someone is manufacturing things that use them. I'll also assume that their stock situation was probably pretty good, simply due to volume.
Given that the audio industry is but a small fraction of total sales, previous remarks about customer service and the target applications are no surprise at all. In that world, saying you're Charles Hansen or John Curl is less impressive to them than the nameless person they deal with from the larger power supply firms. Just imagine how many parts APC alone would go through!
In this industry, anything we have to use in the way of parts came about due to industrial requirements. This industry is, how shall we say ... "high maintenance". That's probably putting it nicely. The people making those cheap LED flashlights are worth more in sales than audio is. We are way, way down in the pecking order.
-Chris
It's just that IR made many parts at affordable prices. A strong motivation when someone is manufacturing things that use them. I'll also assume that their stock situation was probably pretty good, simply due to volume.
Given that the audio industry is but a small fraction of total sales, previous remarks about customer service and the target applications are no surprise at all. In that world, saying you're Charles Hansen or John Curl is less impressive to them than the nameless person they deal with from the larger power supply firms. Just imagine how many parts APC alone would go through!
In this industry, anything we have to use in the way of parts came about due to industrial requirements. This industry is, how shall we say ... "high maintenance". That's probably putting it nicely. The people making those cheap LED flashlights are worth more in sales than audio is. We are way, way down in the pecking order.
-Chris
I think that Mr. Hansen or Mr. Pass don't made products, where they should look at prices of basic devices...I think, that discussion was about behavior of devices and about searching the best results of sound....I know, that all consumer electronic is about only a few percent of electronic market, but it is another question...and I'm not IXYS's dealer, only fan of precise sound...
And, BTW, price of one 500W IXYS is approximately the same like three 150W IR...
And, BTW, price of one 500W IXYS is approximately the same like three 150W IR...
Last edited:
Hi Charles,
One very important thing that you are overlooking in laterals is that they are quite slow (this actually saves some designers from blowing up their amplifers). The problem is that the structure of the lateral inevitably results in fairly high resistance in the gate structure. Few lateral datasheets show it, but if you look at a plot of gm vs. frequency at a given current, you'll see a fairly low frequency cutoff. I believe I mention this in my book.
For those who know how to use verticals and get the most out of them, laterals are a waste of extra money. This is not to say you can't make a good amplifier out of them, of course.
Cheers,
Bob
Hello Torsten, I was looking around the AMR site.
Did you design these?
Abbingdon Music Research - Products - Gold Fuse
Did you design these?
Abbingdon Music Research - Products - Gold Fuse
I tested the Vishay parts, and andy_c pointed out that the data sheet says Vishay Siliconix so perhaps they're made on a Siliconix process rather than IR. There were also comments in the news reports to the effect that Vishay did not purchase the entire IR product line.
I tested 10 parts and they all passed.
There is a caveat...
If you want to run the output stage open loop (no CFP also), you'll need to parallel at least 8 of the Semelab single die parts per polarity. Biasing them each at 400mA will lead you to extreme dissipation values.
I was thinking, if we combine your $10k AMR tube based universal RIAA Re-equalizer PLUS the Myrtle blocks to lift the cables off the floor and then, wait for it, PLUS the gold fuses WOW that would be some vynyl (sic) system. What do you think?
Your best buddy
Job for Ed, quantify the microphonics of various fuses.
Hi,
Not sure I agree.
One pair of the Renesas/Hitatchi parts, open loop, has around 1.12A/V transconductance at quiescent and 0.94A/V if one FET is pushed all the way (10V Vds, 8.5A Id and 9V Vgs).
So one pair has around 1 Ohm output impedance, comparing favourably with many open loop tube Amp's, with +/-55V it will dissipate 44W, delivers the "first watt" in Class A (and should not cause major trouble as it slides into Class B), puts out a bit over 100W into 8 Ohm and can handle around 5 Ohm Load (resistive).
I do not see why 8 pairs are needed.
Ciao T
Not sure I agree.
One pair of the Renesas/Hitatchi parts, open loop, has around 1.12A/V transconductance at quiescent and 0.94A/V if one FET is pushed all the way (10V Vds, 8.5A Id and 9V Vgs).
So one pair has around 1 Ohm output impedance, comparing favourably with many open loop tube Amp's, with +/-55V it will dissipate 44W, delivers the "first watt" in Class A (and should not cause major trouble as it slides into Class B), puts out a bit over 100W into 8 Ohm and can handle around 5 Ohm Load (resistive).
I do not see why 8 pairs are needed.
Ciao T
- Status
- Not open for further replies.