janneman said:
You need to put a diode in series with the collector of Q1, otherwise when Q2 is driven to cut-off when the PNP of the output stage takes over, the B-C (diode) junction of Q1 will become forward biased and conduct. Ditto for the protection transistor in the PNP half.
Cheers,
Glen
A triple-slope derivative of this arrangement.
You may find it necessary to establish the accuracy of your Zener references with changes in current, and any effect this may have on precision and accuracy.
You may find it necessary to establish the accuracy of your Zener references with changes in current, and any effect this may have on precision and accuracy.
mikeks said:
Hi Mike,
Thanks for the doc, I'll read it tonight.
Indeed the zener accuracy is an important point. You will see in the spreadsheet one zener indicated as 12.00001 volt. That is because I used the Exel solver, after the initial calculation, to adjust the R3-R4 ratio until, Z1 = 12V. That helps to at least get one standard value, and 5% zeners are more or less standard nowadays. Also I think that a genaral accuracy of say 5% would be OK for this type of circuits as the SOA and load value and load angle are also rather approximations. The important thing is to err on the safe side.
Another issue is the soft turn-on of the protection transistor. This is of course directly related to the current drive capability of the Vas or driver, depending on the exact topology. What are your experiences with that?
Jan Didden
janneman said:
I am not sure what you mean by "...the soft turn-on of the protection transistor..."
But "better controlled" activation of said device is obtained by inserting a base ballast resistor.
Zener diodes only maintain their nominal tolerances if the current through them is assumed not to vary. This is not the case with that arrangement.
mikeks said:
To the zeners: as you will see in the graph, there is a spice graph. That is the data from an actual simulation put in the same graph. To my own surprise the breakpoints are really quite predictable and stable, with these particular zener models. I have more graphs, for instance from the current through the zeners, I see if i can show them tonight.
Yes, I meant better controlled, although I like soft turn-on because that is what actually happens: the prot transistor turns on slowly when Vbe starts to go from say 500mV to 650mV.
Jan Didden
Hi Janneman,
permit a few comments on the graphical results.
The SOA appear to be for Tc=25degC.
The 8ohm 45degree phase angle load is way outside the DC SOA and getting quite close to the 100mS SOA. This before taking account of Tc de-rating.
The knee @ 20Vce in the Icprot allows more current to pass at low Vce without triggering the protection. But what is the benefit? The 8ohm load lines never approach that area of the SOAR. A dead short into 0r5 will certainly get into that region but by then audibility of the protection is not an issue.
The load line for the reactive load is specified as 8ohm and maybe 45degrees (if I read the variables correctly) but it "looks" more like 60degrees.
Your SOA lines do not appear to mimic the published SOAs for real devices. Have you tried using the log/log display to generate the SOAs? and then display on the linear graph that you prefer.
Can you explain why the Res load-1 or-2 curves are so different?
permit a few comments on the graphical results.
The SOA appear to be for Tc=25degC.
The 8ohm 45degree phase angle load is way outside the DC SOA and getting quite close to the 100mS SOA. This before taking account of Tc de-rating.
The knee @ 20Vce in the Icprot allows more current to pass at low Vce without triggering the protection. But what is the benefit? The 8ohm load lines never approach that area of the SOAR. A dead short into 0r5 will certainly get into that region but by then audibility of the protection is not an issue.
The load line for the reactive load is specified as 8ohm and maybe 45degrees (if I read the variables correctly) but it "looks" more like 60degrees.
Your SOA lines do not appear to mimic the published SOAs for real devices. Have you tried using the log/log display to generate the SOAs? and then display on the linear graph that you prefer.
Can you explain why the Res load-1 or-2 curves are so different?
AndrewT said:
Andrew:
1 - Yes, it's the 25 degr curves from the data sheet.
2 - Yes, I already noted that in my answer to Glen. This is my building of the spreadsheet calculations, not yet a finished design. In fact, what you said indicates that I need two pairs of output devices for this load.
3 - Good point. I think the same point was made by Mikeks in his EW article. I want to reposition the breakpoints to decrease this area and trying to get more area at higher Vce. One of my purposes with this spreadsheet was to be able to 'play' with this.
4 - I will check this. I have since modified the spreadsheet (this was an older version I happened to have at work today).
5 - The SOA is from the SAP16 devices. The problem here is that the data sheet for these devices is rather limited. There is only a single SOA graph and it is not exactly clear what the circumstances and conditions are.
6 - The Res 1 and Res 2 are different resistive load, with onlly res-1 with phase angle added.
Aa said, this is not yet a finished product. What I want to do is this. First, calculate the component values from the initial data as shown on the page with the graph. Then, use those component values and calculate the protection locus. This is the first check: the calculated locus should be the same as the breakpoints initially given; because the calculation is done differently, that's a good check. Finally, I plug the values in the sim, and then use the sim data to again plot the locus. As seen in the graph, the sim is not yet ok especially in the high Vce area.
I feel I need to do this because it is pretty difficult to verify the prot circuit in real life.
Jan Didden
exactly where I am and have been for months.
I am convinced that the model is pretty good, but I cannot find the courage to test that the protection works just like the model says. I fear it may become a test to destruction.
I have seen references to two methods, but I need to calibrate my scope (without a manual) and then soak in some Valium.
Hi,
ESP and Quad (Baxandall), they may be similar. I have still to look for the Quad method.
Both rely on short spikes that generate very little energy in the output stage and load.
The scope picks up the height of the current/voltage spike. I seem to recall Self or Pass mentioning something similar, but not in relation to testing output current. Was it Slew rate limiting?
ESP and Quad (Baxandall), they may be similar. I have still to look for the Quad method.
Both rely on short spikes that generate very little energy in the output stage and load.
The scope picks up the height of the current/voltage spike. I seem to recall Self or Pass mentioning something similar, but not in relation to testing output current. Was it Slew rate limiting?
Hi,
that Stereophile impedance & phase plot shows a minimum Z~=3.6ohms and max phase ~=70degrees.
But these worst conditions do not co-incide, fortunately.
As they point out in the text, the worst combination could be 4.36ohms and 50degrees, still pretty near severe for what may well be specified as a 6ohm speaker.
that Stereophile impedance & phase plot shows a minimum Z~=3.6ohms and max phase ~=70degrees.
But these worst conditions do not co-incide, fortunately.
As they point out in the text, the worst combination could be 4.36ohms and 50degrees, still pretty near severe for what may well be specified as a 6ohm speaker.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.