Did a zoom into the simu – shows that the this discontinuity should *not* be at exactly zero but rather "upstream" than "downstream" (as in measurement)
10W power injection
Would indicate that simu has "room for improvement"
Maybe the relative large area and mass of the attached metal diaphragm of this compression driver plays a role I have overlooked by now
Michael
An externally hosted image should be here but it was not working when we last tested it.
10W power injection
Would indicate that simu has "room for improvement"
Maybe the relative large area and mass of the attached metal diaphragm of this compression driver plays a role I have overlooked by now
Michael
mige0 said:
That would be my guess. "Better" for this purpose anyway. Assuming a typical solid state class AB amp, it's not hard to imagine the bias compensation mistracking the output device thermal properties - leading to strange crossover distortion. Note the discussion over on the solid state thread regarding Thermaltrak devices. I wouldn't think that a 10W sine distortion test at 3Hz is part of the design requirements of a typical home audio amp.
The same test done with a coil of the DCR, but much higher thermal mass would isolate the amp issue.
Sheldon
Sheldon said:
Yes was my first guess too - but I checked back the voltage measurement over the VC taken simultaneously, an this trace isn't distorted at all.
Maybe the HP of the soundcard input could do that kind of shift ?
I'm thinking of the squarewave shape deformation by high pass filters here.
Michael
mige0 said:
Silly me
– what is SPICE good for if not to simulate exactly this !!An externally hosted image should be here but it was not working when we last tested it.
10W power injection - exactly the same zoomed simu as shown before in posting #181
http://www.diyaudio.com/forums/showthread.php?postid=1686394#post1686394
but "seen through" a HP of roughly 2Hz.
Haven't checked the input of my soundcard yet - what HP they *exactly* use .
Though "Willy's fin" still seems to look different – maybe my soundcards high pass input is of higher order – have to check it out.
Michael
gedlee said:
Yeah - possibly...
But what exactly do you mean with "not symmetric" heating and cooling.
I have outlined it in detail and John has said the same in words that may be closer to "your speach".
There is a mean value - tending asymptotically towards a certain equilibrium value.
And there is the "thermal overshot" swing around this mean value.
I don't see anything else - and think you have another mental hurdle here.
Can you elaborate on your point – maybe I misunderstood?
Michael
PS :
Did you got around the HP filtered file?
Pan said:
Cables ? - don't think so – hadn't have any cross talk issues until now.
I mostly use my own cables as standard confection make XLR's – and even good and expensive PRO ones, I ordered on my specifications – have shown issues.
Also speaker cable and sensing cables go opposite directions from DUT. No – almost impossible.
Yes – my shunt used on this measurement is *extreme* low - around 5mOhm and connected in four wire technique - otherwise you immediately see the current envelope shape in the voltage measurement across DUT (keeping things isolated from each other is one of the huge benefits of using symmetric inputs
). With this shunt I have a comfortable 40dB isolation between voltage and current measurement – way enough for visual checking of wave shape.You'll need a very good (mic-) pre to do clean measurements on such low signal levels though!
Michael
mige0 said:
Should have been:
"a comfortable 80dB isolation"...
---------
Thanks for your help guys - but we all went wrong – me, (shame on me) who could have known better - *the most* of all...
Its always good you have full documentation of the equipment used
– even better though, when you use the documentation you have ! Basically two (the ones I mainly used until now) out of three of the inputs of my soundcard look like that:
An externally hosted image should be here but it was not working when we last tested it.
What we see is the input network of a (excellent) symmetric mic-preamp that also can be plugged at line level (at a higher input impedance) .
Sadly it hasn't exactly been designed for extreme low frequency inputs (overlaid by DC voltage).
Simply put – the coupled input choke can't handle 3Hz signals as distortion free as I would need it for my measurements.
Hence my measurements on the level of wave form deformation were misleading and my attempt to get rev 2.3 model into alignment with these measurements resulted in a buggy model.
You could say my rev 2.3 model was "pretty good" – to mimic the *wrong* measurements
(to somehow make the best out of it)
consequence :
> lets trash rev. 2.3 thermal model at this point
> lets go back to rev 2.0 model
> lets continue developing towards rev 2.4 model
Will come back with my next attempt on bridging the gap of quantifying thermal behaviour on long / mid term TD *and* TD on the level of wave form deformation.
Also have to sort out what wasn't valid qualitatively or quantitatively, presented earlier.
First superficial survey: Thermal Distortion on the level of wave form deformation was affected the most – especially regarding the *time domain plots* I showed from simus and the Tungsten bulb measurement.
FFT plots from measurements and simus shift in relative amplitude but not really qualitatively seen.
SPL Melt down plots with steady sine signal and burst sine are not affected – sorry Earl
- bad news for your asymmetric heating / cooling theory you won't like to explain to us.Looking at the bright side of things -
– I'm quite happy having found the reason to trash my rev 2.3 model, as it caused me a lot of headache with - what it appeared to me - its un-logical, un-intuitively and un-elegant translation of physics into math.
Michael
Rev 3.1.2 thermal model is ready and again there is free download at
http://members.aon.at/kinotechnik/diyaudio/diy_audio/TD/SPICE_modelling_TD.htm
After the steep set back with my last model I had to
- first upgrade my measurement set-up by DC instrument pre-amplifiers
- add some more blocks to compensate for measurement equipment set-up in SPICE
- completely re-arrange the loop modelling constant voltage / constant current amp operation
- do a lot of measurements and simus to get a clear idea of what exactly went wrong
The outcome is a model that closes the gap of Thermal Distortion from long term behaviour down to wave form deformation.
Also Rev 3.1.2 model provides
- ready for use modelling of high pass filter of amplifier
- ready for use modelling of source impedance of amplifier, cables and connectors
- ready for use modelling of high pass filter of sound card input
- Modelled speaker impedance response can be inserted as a series network
and keeps the known features
- Temperature increase markers for Safe Operating Area are provided
- Single sinus and dual sinus signal generation ready for use
- Burst on/off generation ready for use to simulate Thermal Overshot effects
- cooling due to radiation
- expandability of the core thermal model for even more detailed thermal modelling
- parameter entry and normalised signal output
- a structured layout, meaningful parameter names and a brief HowTo
It now is possible to simulate pure constant voltage amplifier operation or pure constant current amplifier operation or anything in between.
SOA analysis looks different now as resistance increase and SPL melt down are completely different for current or voltage sources.
With current sources there is – of course – temperature increase, but there is *no* SPL melt down as voltage over DUT increases to compensate for resistance increase due to heat up –keeping the current through DUT constant.
Deeper investigation on the failure of last model showed that there is no distortion of the Mackie Onyx mic-pre-amp involved (at least not in the sense of "non-linear distortion")
The coupled choke at the mic input is OK !
All hassles came from an unexpected shift in high pass frequency *and* in a shift of the *order* of the high pass slope when changing the adjustable gain of these inputs.
Best advice for measurement is to use high gain / DC coupled instrument pre amps (SSM2019, AMP02, INA217, THAT1510 ...) plugged to a (not adjustable) line in.
Tapping down the "voltage across DUT" signal *to the same value* of the shunt voltage by resistance divider trimming ensures absolutely the same behaviour for both, the current and the voltage measurement (make sure this divider is of high enough impedance to stay cool).
Using the gain adjustment feature of the DC instrument pre amps rather than the unpredictable gain adjustment of the soundcard does not affect critical HP slope and keeps HP frequency constant and as low as possible – greatly simplifying measurements *and* necessary compensation for simulation.
Above is how wave form deformation from Thermal Distortion can look like – simulated by Rev 3.1.2 model.
Playing around with the soundcard input HP filter provided by Rev 3.1.2 model to mimic the generic Mackie Onyx mic inputs results in the "not valid" stamped time domain plots shown earlier.
It simply was "linear distortion" – not even distortion at all where I come from.
No big deal – but extremely confusing to me.
The question about higher odd order Thermal Distortion (5th 7th ...) is also clarified by the new model. Yes they are there - but they are very, very low.
Wish you all a happy new year
(and hopefully *no fix* (!) of the so called "crisis" - which in fact is inherent part of the long term cycle - but rather a micro cosmic change in thinking for the better) !
Michael
http://members.aon.at/kinotechnik/diyaudio/diy_audio/TD/SPICE_modelling_TD.htm
An externally hosted image should be here but it was not working when we last tested it.
After the steep set back with my last model I had to
- first upgrade my measurement set-up by DC instrument pre-amplifiers
- add some more blocks to compensate for measurement equipment set-up in SPICE
- completely re-arrange the loop modelling constant voltage / constant current amp operation
- do a lot of measurements and simus to get a clear idea of what exactly went wrong
The outcome is a model that closes the gap of Thermal Distortion from long term behaviour down to wave form deformation.
Also Rev 3.1.2 model provides
- ready for use modelling of high pass filter of amplifier
- ready for use modelling of source impedance of amplifier, cables and connectors
- ready for use modelling of high pass filter of sound card input
- Modelled speaker impedance response can be inserted as a series network
and keeps the known features
- Temperature increase markers for Safe Operating Area are provided
- Single sinus and dual sinus signal generation ready for use
- Burst on/off generation ready for use to simulate Thermal Overshot effects
- cooling due to radiation
- expandability of the core thermal model for even more detailed thermal modelling
- parameter entry and normalised signal output
- a structured layout, meaningful parameter names and a brief HowTo
It now is possible to simulate pure constant voltage amplifier operation or pure constant current amplifier operation or anything in between.
SOA analysis looks different now as resistance increase and SPL melt down are completely different for current or voltage sources.
With current sources there is – of course – temperature increase, but there is *no* SPL melt down as voltage over DUT increases to compensate for resistance increase due to heat up –keeping the current through DUT constant.
Deeper investigation on the failure of last model showed that there is no distortion of the Mackie Onyx mic-pre-amp involved (at least not in the sense of "non-linear distortion")
The coupled choke at the mic input is OK !
All hassles came from an unexpected shift in high pass frequency *and* in a shift of the *order* of the high pass slope when changing the adjustable gain of these inputs.
Best advice for measurement is to use high gain / DC coupled instrument pre amps (SSM2019, AMP02, INA217, THAT1510 ...) plugged to a (not adjustable) line in.
Tapping down the "voltage across DUT" signal *to the same value* of the shunt voltage by resistance divider trimming ensures absolutely the same behaviour for both, the current and the voltage measurement (make sure this divider is of high enough impedance to stay cool).
Using the gain adjustment feature of the DC instrument pre amps rather than the unpredictable gain adjustment of the soundcard does not affect critical HP slope and keeps HP frequency constant and as low as possible – greatly simplifying measurements *and* necessary compensation for simulation.
An externally hosted image should be here but it was not working when we last tested it.
Above is how wave form deformation from Thermal Distortion can look like – simulated by Rev 3.1.2 model.
Playing around with the soundcard input HP filter provided by Rev 3.1.2 model to mimic the generic Mackie Onyx mic inputs results in the "not valid" stamped time domain plots shown earlier.
It simply was "linear distortion" – not even distortion at all where I come from.
No big deal – but extremely confusing to me.
The question about higher odd order Thermal Distortion (5th 7th ...) is also clarified by the new model. Yes they are there - but they are very, very low.
Wish you all a happy new year
(and hopefully *no fix* (!) of the so called "crisis" - which in fact is inherent part of the long term cycle - but rather a micro cosmic change in thinking for the better) !
Michael
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.