Assembly time is five minutes.Direct replacement with BYV 95A soft recovery 1A diodes.
Attachments
Listening test with a difference.
I've been having a play with three different power amplifiers and experimenting with capturing the audio across the speaker output.
This one is very very different to anything we have done before 😉
http://www.diyaudio.com/forums/lounge/303727-welcome-virtual-listening-room.html
I've been having a play with three different power amplifiers and experimenting with capturing the audio across the speaker output.
This one is very very different to anything we have done before 😉
http://www.diyaudio.com/forums/lounge/303727-welcome-virtual-listening-room.html
Can I bring up another topic? Safe Operating Area of power output devices.
I don't want to rehash all the details, we know what SOA means and why it can be a source of smoke and smell. 🙂
My question is this: if I make sure (by whatever means) that the junction temperature of the devices remains within safe limits, does that then mean there is no danger of hot spots developing and secondary breakdown occuring?
Jan
I don't want to rehash all the details, we know what SOA means and why it can be a source of smoke and smell. 🙂
My question is this: if I make sure (by whatever means) that the junction temperature of the devices remains within safe limits, does that then mean there is no danger of hot spots developing and secondary breakdown occuring?
Jan
In my design, this is the case:
"MOSFET Thermal runaway in linear mode.
[snip]Recently, MOSFETs with very high transconductance, optimised for switching operation, have become available. When operated in linear mode, especially at high drain-source voltages and low drain currents, the gate-source voltage tends to be very close to the threshold voltage. Unfortunately the threshold voltage decreases as temperature increases, so that if there are any slight temperature variations across the chip, then the hotter regions will tend to carry more current than the cooler regions when Vgs is very close to Vth. This can lead to thermal runaway and the destruction of the MOSFET even when it is operating within its Vds, Id and Pd ratings."
Let me detail my question: how much delta-T would there be across the die? Say that the max allowed Tj is 150C, and I limit it to say 100C, I have a 'safety margin' of 50C for delta-T across the die as a result of the above phenomenon. Am I safe?
Jan
Do yu have a plot of the threshold vs. temperature curve? On traditional Mosfets the curve works in your favor with the threshold voltage going up with temperature. The HV mosfets you are looking at may be quite different. They are suggesting that only a few degrees difference would cause issues. A curve tracer and heat probe to warm up the device may be the only way to a real answer. Which parts are you looking at?
Its one of IXYS HV MOSFETs, see attached. I don't think there is a threshold vs. T curve. In my application, the (linear) Id is just a dozen mA or so, which means that the Vth varies very little as described in my above post. Vds does go up close to the absolute 4500V limit however.
I have also corresponded with IXYS support about the curious shape of the SOA graphs, but that didn't really lead to anything ('we don't use them in linear mode and we don't have the right test equipment to do those SOA curves').
I have a real-time hardware solution to estimate Tj which should be pretty accurate, so I can monitor that and make sure it doesn't go over a specific limit. I would hope I'd be done with that...
Edit: I notice you mentioned Vth to go up with T, but the quoted piece from Wikipedia says Vth drops with T...
Jan
Last edited:
This is my thinking, hot-spots are the most common source of second-breakdown, so if this thermal run-away you are talking about is the source of the 'second/upper' knee/bend in the SOA-graph then you should be safe if you keep things below that 'second/upper' knee/bend in that graph (but I could be wrong).
Edit: I was going to say that this would be easy to (destructively) test, but at near 40 Euro ... 🙂
Edit: I was going to say that this would be easy to (destructively) test, but at near 40 Euro ... 🙂
Last edited:
Well the MOSFETs don't have a 2nd breakdown as a BJT has, but apparent there *could* be a similar effect from the Vth temp dependence at very low currents. The jury is still out however.
At say 10mA Id in a 1A device there doesn't appear much to be hogged anyway.
Jan
At say 10mA Id in a 1A device there doesn't appear much to be hogged anyway.
Jan
We made many current hogging IR videos of DSL drivers. One customer was blowing holes in the package with the explosive failures
.
As I recall, the IR line had a current where the Tc was zero. Operation above was negative coeff, below was positive coeff. There was an article in one of the IR app books that detailed that for linear and switching operation.
Just searched after I wrote that, found this:
http://www.irf.com/product-info/datasheets/data/irfm450.pdf
Page 3 figure 3. Zero tc is at 6.25 amperes with 50 volts.
John
High voltage amps are a real specialty. AC calibrators use them but direct amps seem to have fallen to history and the current generation use transformers. The older designs used tube/transistor hybrids. The current Fluke design is here :http://download.flukecal.com/pub/literature/5725a__imeng0700.pdf with complete schematics but its internal output is 400V to the transformers and they use 4 transformers and 5 connections to get from 10 Hz to 100 KHz.. The older Fluke and HP designs are on the web. The current Fluke does use FET's for the voltage amp and output stages. Tons of safety issues come into play as well.
Most of the other stuff I have found makes it to 1 KHz.
Are you using 1 transistor for the full 4K swing? No totem pole?
Most of the other stuff I have found makes it to 1 KHz.
Are you using 1 transistor for the full 4K swing? No totem pole?
You want this. Unfortunately it is still too vague for accurate estimation
of the problem with the big Mosfets designed for switching, that is to say
you apparently cannot trust even the DC curve for linear operation in this
regard. Ixsys does have some devices rated for safe operation in this
regard, achieved through special layout and some resistive ballasting locally
on the chip.
of the problem with the big Mosfets designed for switching, that is to say
you apparently cannot trust even the DC curve for linear operation in this
regard. Ixsys does have some devices rated for safe operation in this
regard, achieved through special layout and some resistive ballasting locally
on the chip.
Actually the secondary breakdown area is to the right of the 'second/right' knee in the graph (not the 'second/upper' knee. Sorry for the slip-op-pen).
https://en.wikipedia.org/wiki/Safe_operating_area
In my opinion, the graph in this datasheet shows that secondary breakdown knee
http://www.diyaudio.com/forums/atta...torch-preamplifier-part-ii-ixtt-th1n450hv.pdf
P.s. That datasheet does not offer an other explanation for that area of the SOA curve, this is why I did describe it as secondary breakdown area (and yes I do know that Mosfet's do not exhibit 'classical/BJT-like' secondary breakdown).
Last edited:
i observed that the temperature stability of mosfets differ markedly across process and materials.
I found (as a crude generalization) that high Gm mosfets can be tough to tame and that Sics are also quite erratic, despite their low Gm and high temperature capabilities.
I found (as a crude generalization) that high Gm mosfets can be tough to tame and that Sics are also quite erratic, despite their low Gm and high temperature capabilities.
Thanks all for the pointers, lots of stuff to study, it'll take me a day or so.
To be sure, the amp is working fine and so far no issues, but before getting it on the road I really would like a sweet and fuzzy feeling that I actually know all that is going on inside 😉
Jan
To be sure, the amp is working fine and so far no issues, but before getting it on the road I really would like a sweet and fuzzy feeling that I actually know all that is going on inside 😉
Jan
John,
There is no specific graph in the 1N450 data sheet, but figs 1 & 2 seem to hold the information. If I do a quick eye-ball acquisition it seems that the Vgs vs Id tempco is negative for Vgs = 6V, and positive for Vgs = 7V. Unfortunately, in my application Id gets up to max 120mA or so in class AB, which is in the Vgs = 6V area.
Not trying to 2nd guess the data sheet (or maybe I am) but this is a very low Ids max and maybe I am lucky? Standing DC current is about 15mA max at Vds = 2.1kV.
Can I test this by looking at the leakage current at max Vds with T as a variable?
Jan
- Status
- Not open for further replies.