Hi Ilimzn,
your post, as usual, is very informative and technically impeccable
Exemplary.
But I take exception to your introduction:
But it gets worse
I do not have an intimate knowledge of the power semiconductor market, in fact, it is very limited, to just a few manufacturers and even fewer of their devices.
I can however, recall an ONsemi BJT device, namely MJ21193/4 that is more than a match for the irfp240.
The BJT does indeed suffer secondary breakdown. Look at the 200Vdc ratings, the 21193 can pass 600mA and 240 does 750mA, pretty close to “wish for”.
However, at the typical voltages that the quasi will run at, about +-70Vdc, (and cold temperature Tc=25degC) the 21193 can pass 3571mA and 240 only 2714mA. More than one could “wish for”.
Taking account of temperature derating, as you rightly pointed out, allows the 21193 to draw even further ahead. As a result of the higher Tcmax (200degC) of the 21193 and at typical strenuous operating temperatures of say Tc<=90degC, the 21193 rating is 2245mA @ 70Vdc. The p240 can pass 1029mA @ 70Vdc and the same limiting operating temperature.
It seems to me that the FET is outclassed by the BJT when medium voltage and high temperature operating conditions are applied.
Beginning Electronics learners are more likely to remember your conclusions and not the substance of the argument.
It would be wrong to leave them with a misleading conclusion that is patently untrue in some senarios.
Comments like your introduction are beneath you.
I know you can do better, I have seen it here and in many other posts.
your post, as usual, is very informative and technically impeccable
Well argued and even comments added pointing out further investigation of the pitfalls of selecting alternative manufacturer's components. The consequence of different specifications is aired and shows that re-designing/checking is necessary for even apparently same component replacement.ilimzn said:
Exemplary.
But I take exception to your introduction:
by stating that FETs do not have secondary breakdown implies that by using FETs one can shortcut the design process. This is misleading. The full SOAR calculations and appropriate derating for operational temperatures still need to be carried out for both BJT or FET output devices.ilimzn said:
But it gets worse
is completely untrue.ilimzn said:
I do not have an intimate knowledge of the power semiconductor market, in fact, it is very limited, to just a few manufacturers and even fewer of their devices.
I can however, recall an ONsemi BJT device, namely MJ21193/4 that is more than a match for the irfp240.
The BJT does indeed suffer secondary breakdown. Look at the 200Vdc ratings, the 21193 can pass 600mA and 240 does 750mA, pretty close to “wish for”.
However, at the typical voltages that the quasi will run at, about +-70Vdc, (and cold temperature Tc=25degC) the 21193 can pass 3571mA and 240 only 2714mA. More than one could “wish for”.
Taking account of temperature derating, as you rightly pointed out, allows the 21193 to draw even further ahead. As a result of the higher Tcmax (200degC) of the 21193 and at typical strenuous operating temperatures of say Tc<=90degC, the 21193 rating is 2245mA @ 70Vdc. The p240 can pass 1029mA @ 70Vdc and the same limiting operating temperature.
It seems to me that the FET is outclassed by the BJT when medium voltage and high temperature operating conditions are applied.
Beginning Electronics learners are more likely to remember your conclusions and not the substance of the argument.
It would be wrong to leave them with a misleading conclusion that is patently untrue in some senarios.
Comments like your introduction are beneath you.
I know you can do better, I have seen it here and in many other posts.
AndrewT said:
i can only agree - and in fact, didmn't i write:
'Since MOSFETs do not have secondary breakdown problems, SOAR is calculated only from the dissipation and it's derating, i.e. what kind of heatsink you have and how well you can thermally interface the MOSFET case to it.'
...i.e, everything you would usually need, except for secondary breakdown considerations. These are, in fact, what makes SOAR calculations quite difficult for BJTs.
While I find your math impeccable, you have neglected one VERY important parameter: MJ21193/4 have a maximum dissipation of 250W, versus 150W for the IRFP240.
It should be noted that when we say 'maximum power dissipation' we are in fact providing a maximum limit on the SOAR with no other effects to account for.
I did not explicitly say it, because I would expect it to be obvious that I was not comparing the SOAR of the IRFP240 to ANY avaialble bipolar device - just to ones with equal maximum power dissipation, which is how one usually compares SOAR curves.
Unfortunately, there is no completely equivalent device we can compare to, but something from the MJW series with up to 150W dissipation is a good start. To make things fair, we can look at the part of the SOAR curve below 15A (since that is the limit for MJW parts), and 200V (since that is the limit for IRFP240). What you find is simple: secondary breakdown limits the power dissipated at higher voltages across CE for the BJT - and in fact, this is quite signifficant, do not foget t hat the diagram uses exponential scale.
To sum it up:
On the same Pd basis, MOSFETS willalways have better SOAR than BJTs, because BJTs miss that part of the SOAR surface chopped off by secondary breakdown. Not having secondary breakdown is what BJTs can only wish for.
Also, because Vce max is often increased to improve secondary breakdown characteristics, for said improved characteristiscs, BJTs tend to have lower maximum current. This makes MOSFETs a little more flexible, because you can chose between almost any combination of voltage and current for a given Pd limit, without affecting SOAR. However, MOSFETs pay for this by having substantially worse P channel parts, compared to N channel.
ilimzn said:
I'd love to catch Ilimzn on a hasty statement one day.
Unfortunately the one above looks airtight, Again.
I have the feeling a great many do not really have a grasp of what secondary (thermal) breakdown is, and the mechanism of the oxydation layer of Mosfets that enables these devices to lack this feature.
Why 1+(gm * Rload)?
I think, 1+Cgs*K, where K is actual D-S swing/S-G swing, and it is the product of feedbacks, instead of gm and Rload.
That's why I prefer to use power FETs in source followers, because in such case input capacitance is predominantly Cgd + Cgs*(1-K), where K is close to 1).
Anyway the amp is asymmetrical, so I'd use MOSFET in the + rail, and NPN BJT in the - rail (why NPN - because emitter followers working on complex loads distort more by definition).
I think, 1+Cgs*K, where K is actual D-S swing/S-G swing, and it is the product of feedbacks, instead of gm and Rload.
That's why I prefer to use power FETs in source followers, because in such case input capacitance is predominantly Cgd + Cgs*(1-K), where K is close to 1).
Anyway the amp is asymmetrical, so I'd use MOSFET in the + rail, and NPN BJT in the - rail (why NPN - because emitter followers working on complex loads distort more by definition).
Hi Ilimzn,
let's accept your new rules, Yes, I will allow you to change midstream.
Take the figures I quoted earlier for the p240 and factor them up to the hypothetical 250W.
The 250W FET passes 1715mA @ 70Vdc and <=90degC. instead of 1029mA.
That is still well short of the To3 cased 21193/4 at 2245mA at the same limiting conditions.
I can't believe that the performance of the 21193/4 is unique, I suspect there are other devices that can similarly outperform FETs in real world audio amplifier operating conditions.
I do not take back my objection to your all encompassing conclusion that still misleads the ill-informed of the true comparison between FETs and BJTs.
Edit
the 190W p450 can only manage 1303mA for the comparable conditions.
let's accept your new rules, Yes, I will allow you to change midstream.
Now, let's invent a hypothetical FET that has 250W dissipation to make it comparable to that nice cheap and available MJ21193/4 BJT.
Take the figures I quoted earlier for the p240 and factor them up to the hypothetical 250W.
The 250W FET passes 1715mA @ 70Vdc and <=90degC. instead of 1029mA.
That is still well short of the To3 cased 21193/4 at 2245mA at the same limiting conditions.
I can't believe that the performance of the 21193/4 is unique, I suspect there are other devices that can similarly outperform FETs in real world audio amplifier operating conditions.
I do not take back my objection to your all encompassing conclusion that still misleads the ill-informed of the true comparison between FETs and BJTs.
Edit
the 190W p450 can only manage 1303mA for the comparable conditions.
AndrewT said:
Ok, I'll show you a picture where B-E junction looks like a diode. Right, this picture... Looks like a rectifier? Agree... Now add a filter capacitor to it... Right... And a load... Right, this way... However, it is not so good rectifier to power another amp from the output, but anyway it shifts bias on high frequencies... Now, can you explain me what this famous so called "Transistor Sound Phenomena" means? You a right! It is an effect of rectification on high end of the spector by typical transistor amp with an emitter follower!
Now, do you remember my article "Swinik VS Quad"? What it was about? Right! It was about amp designs that despite of lack if idle current that believed to cause the "Transistor Sound Phenomena" don't sound like transistors! Why?
Wavebourn said:
Actually, gm * Rload is close to your D-S swing. This remains unchanged regardless of overal feedback since this is the GS swing the MOSFET requires to pruduce a current swing - with or without feedback. This is important to remember as it introduces a mechanism that makes HF distortion higher with lowering of Rload, lowering the load automatically presents more apparent capacitance to the previous stage by requiring more GS swing for the same SD swing.
Why 1+gm*Rload? Simply because that is the voltage swing Cdg sees - for each V upwards on G, you get gm * Rload volts downwards.
Andrew;
as I told before, this output stage sounds increadible, despite it is totally asymmetrical: one source follower is loaded by a BJT current source controlled by input signal:
Idle current = beta of npn transistor * (one half of powering voltage / resistor in emitter of pnp transistor), and it is equal to half of the current needed for the current peak on the load.
as I told before, this output stage sounds increadible, despite it is totally asymmetrical: one source follower is loaded by a BJT current source controlled by input signal:
Idle current = beta of npn transistor * (one half of powering voltage / resistor in emitter of pnp transistor), and it is equal to half of the current needed for the current peak on the load.
AndrewT said:
How is this changing midstream? You can't compare two devices on SOAR unless it's on equal ground. You think it would be impossible for me to find a comparable 250W Pdiss MOSFET?
The whole point of what I said is, on equal terms, MOSFETs have that part of the SOAR area which is denied the BJT, and that is the secondary breakdown area - something BJTs can only wish for. I don't see what the problem is with that statement, although I obviously have to overstate it with all possible brackets, lest you go comparing a 250W BJT with a 1W MOSFET and claim it has higher SOAR - of course it has. The claim I made was NOT made speciffically to a IRFP240 and I really don't see where I wrote that (maybe it's the fact english is not my native language?) - it's by no means a magical part (although I am aware audio tends to use a lot of parts claiming to be so lately). Besides, I also mentioned IRFP450, why not compare with that as well?
Actually, FYI, it is not as easy to invent a FET this way. A larger dissipation FET implies either a larger die, which in turn implies lower die to case thermal resistance, or a larger case (again lower thermal resistance) or both. In any case, this makes calculations quite different. Just in this case, there is no comparable FET in a TO3 case, signifficant because TO3 has a lower case to heatsink thermal resistance due to a larger interface area. Also, SOAR is given on the ON-semi datasheet for the MJ21193 for 1s, IRFPxxx is given for DC. All in all, you can't scale up only the parameters you like, it's not that simple. Compared on strictly equal ground, the thing you would see with a FET is that there would be no kink at 75V due to second breakdown, and this is the extra SOAR you get - no more, no less. And that is ALL I implied.
I honestly don't see what the hangup is here. If there still is one, I respectfully ask for the forums collective forgiveness and promise I will write all possible caveats with any assertion I make, just so that a member would not assume something I had not written, because of some obviously unjustified expectation of mine that I need not cover the basics all over again in every post. This will no doubt make my posts several times longer than they already are. It will also make me sorry I post at all.
Wavebourn said:
Exactly - which is what I did to get some relative figures, so a comparison can be made. Of course, with the other half of the circuit working, gm changes (exactly how much, depends on bias current!), and, with source resistances, effective gm is lowered even though GD swing is slightly increased, but Cgs is 'bootstrapped' so appears lower.
It should be noted that the top MOSFET in quasi's amp does not work exactly as a follower, due to 'bootstrapped' R between gate and output. In fact, this makes the node impedance higher, as high as the corresponding node on the bottom half of the output. It is curious that it enforces simmetry by actually 'worsening' the situation for the top MOSFET in order to make it act the same as the bottom one
ilimzn said:
Exactly - which is what I did to get some relative figures, so a comparison can be made. Of course, with the other half of the circuit working, gm changes (exactly how much, depends on bias current!), and, with source resistances, effective gm is lowered even though GD swing is slightly increased, but Cgs is 'bootstrapped' so appears lower.
It should be noted that the top MOSFET in quasi's amp does not work exactly as a follower, due to 'bootstrapped' R between gate and output. In fact, this makes the node impedance higher, as high as the corresponding node on the bottom half of the output. It is curious that it enforces simmetry by actually 'worsening' the situation for the top MOSFET in order to make it act the same as the bottom one
Due to a global feedback more current is applied to the gate on higher frequencies, but the resistor actually improves the stage making this current variations less. It helps the lower shoulder as well. A bit. If to make resistors much less it improves performances drammatically, but requires a giant VAS stage, for example see Swinik-III schematics.
Just to add my 2 cents worth, I have in the last year built 2 identical amplifiers bases on my AV400, One with 2sc5200 and 2sa1943 BJTs and the other with IRFP240 and IRFP9240
in the output stage both were tested on 4 Ohm loads @ around 550 watts. Both amplifiers Were thermally very stable and had no Oscillation issues. and both had the same number of o/p devices.
After a number of hours testing both amplifiers at high power, I had the BJT un-expectantly fail in the output stage for no apparent reason. They just shorted out.
It has been my experience over the years that BJTs will not take extented stressing of the o/p stage compered to Hexfets.
I have had my AV1000 be accidently shorted out at full power and the only thing that got damaged was one Output fet went a bit leaky in the Gate.
In practise IMHO Hexfets are much more rugged devices than BJTs.
While at the same time some of the most powerful amplifiers built use BJTs in the Output stage and are very rugged.
It just shows in this this case it wasn't the case.
in the output stage both were tested on 4 Ohm loads @ around 550 watts. Both amplifiers Were thermally very stable and had no Oscillation issues. and both had the same number of o/p devices.
After a number of hours testing both amplifiers at high power, I had the BJT un-expectantly fail in the output stage for no apparent reason. They just shorted out.
It has been my experience over the years that BJTs will not take extented stressing of the o/p stage compered to Hexfets.
I have had my AV1000 be accidently shorted out at full power and the only thing that got damaged was one Output fet went a bit leaky in the Gate.
In practise IMHO Hexfets are much more rugged devices than BJTs.
While at the same time some of the most powerful amplifiers built use BJTs in the Output stage and are very rugged.
It just shows in this this case it wasn't the case.
Hi Ilimzn,
All I ask is that you don't "put down" BJTs in that way.
Experienced designers already know that FETs and BJTs behave differently at high Vce.
In experienced/ beginning designers simply need reminding to go and check the SOA of the device being adopted. No more is required, unless they ask "how do I check".
All I ask is that you don't "put down" BJTs in that way.
Experienced designers already know that FETs and BJTs behave differently at high Vce.
In experienced/ beginning designers simply need reminding to go and check the SOA of the device being adopted. No more is required, unless they ask "how do I check".
I did in my edit, 1303mA cf 2245mA.
Hi Saint,
I am inclined to agree with you.
In general SS output stages need to be designed for the duty they have to perform.
From little experience I can see from the numbers you quoted (550W into 4r) that the 5200/1943 is a bad choice for that amplifier.
That device has an atrocious high Vce performance and shows up well the difference that secondary breakdown can have on longevity when the design is not thorough or simply incorrectly applied.
I am inclined to agree with you.
In general SS output stages need to be designed for the duty they have to perform.
From little experience I can see from the numbers you quoted (550W into 4r) that the 5200/1943 is a bad choice for that amplifier.
That device has an atrocious high Vce performance and shows up well the difference that secondary breakdown can have on longevity when the design is not thorough or simply incorrectly applied.
AndrewT said:
I can only agree with that, besides, I was not 'putting down' BJTs. I use both in my designs, and of course, anyone who does not do their work regarding the SOA of their chosen device, will have it blow in their faces at some point, it's that simple.
Still, in my experience, just like Saint's, HEXFETs tend to be more rugged, although they have their own problems in application, which is something I wrote about earlyer in this thread.
Actually, there's nothing wrong with the devices, it depends how you use them (bridge, how many pairs) and what you want of them. For instance, although MJ21193/4 have exemplary second breakdown performance (in fact, it is quite difficult to find something to compare them with in this regard, outside of the same series of BJTs), they pay for it with lower gain and GBW. The Toshiba parts, OTOH, are 'array transistors' and manage much better in this regard - but pay for it with worse second breakdown performance. A LOT has been done technologically in the last 30 years or so to make this less of a compromise. On the other end of the scale, a MOSFET has fewer inherent speed limits, but transfers the problem to it's driver. Interestingly most array BJTs also show rather high input capacitances, similar to MOSFETs (this is very rarely talked about, strangely enough).
So, to sum it up, it is entirely possible to make versions of an aplifier with many different types and kinds of output devices, if you do your calculations - the big question is what is 'optimal', performance vs pricewise.
Wavebourn said:
Also using current sources would improve performance somewhat, but it sadly leaves you with indeterminate DC conditions. MOSFET amp performance tends to improve when more current is available to charge and discharge the capacitances, the only problem is you end up with driver stages that could almost drive the output themselves if you go too far
The better solution is to find better MOSFETs
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.