heh.
you noticed!
yes, I machined down aluminum hex standoffs to the proper size.
this was the set up for testing and trimming.
that was a stereo amp. about 180watts into 8 ohms, Hitachi Mosfets. this was the MkII version, toroidal xfmrs that iirc were totaling about 1800kva. the original model had stacked core xfmrs that I was unable to find the current limit with and quit around 4kva, figured they were "good enough". so, to review quickly, 1/2F, 1.8kva power supply, 400 amp diodes and a 180 watt amp. 🙂
btw, they were fairly extraordinary bridged - actually scary good.
pretty good SOA, as I ran them into what was essentially a short, the spade lugs touching while the speakers ran at normal volume - I noticed the smell of electronics getting hot (know that smell?) and wondered what it was. went over to the amp and noticed the heatskink on one side was getting warmer than usual. as I was switching off the amp one of the Mosfets on one of the rows decided to arc weld itself through the top of the case!
other than that, the amp was fine and I replaced the blown Mosfet with another matched one (luckily I had a match). btw, no degeneration resistors on the Mosfets.
_-_-
But the only point of my post was to contrast the power supply design philosophies.
I could only find one 400A diode that might call itself HEXFRED: datasheet here. What part number of 400A diodes did you use back in 1991?
thanks I remember those Hitachi's well I still have the data book from the eighties. I built and tested the Erno Borbely power amp using 2 pairs per side from 'Audio Amateur' back in 89 or so.
I don't know what "common concensus" is here?
But physical laws are not easily impressed by common consensus...
And i can not follow your line of thoughts. It seems that impedance is a central star in your thinking 🙂
Have you done any measurements to see the simulated sharp edges?
If the series resistance in the supply is small enough that there is only a small voltage drop due to DC out then you can regard the sum of the C's as the reservoir capacitance. If not, then the reservoir is only the first C. I think in this context 'small voltage drop' means small compared with the ripple, not small compare with the supply rail. The issue is whether at the main ripple frequency the set of caps all work as one or as a reservoir followed by a low pass filter.HiFiNutNut said:
Just to see which ballpark we are in, take 12000uF and 0.5R. This has a corner frequency of around 26Hz, so we are probably (just) in the regime of reservoir followed by low pass filter. So at very low frequencies the effective impedance of the supply is given by half the ripple voltage at the first cap, divided by DC current. Your best bet is to model it with PSUD2 - easier than doing calculations. See how the output voltage varies with current draw.
These were single diodes, back when IR was an independent company.
I think I still have a few somewhere... but offhand I have no idea what the specific part number was. It was however a HEXFRED, they were fairly new at the time. Did have the threaded screw hole on the top as a terminal and a metal base as the other terminal with two mounting holes, one on each side. That side was intended to go to a heatsink. In my situation that was the copper buss bars, but I never drew enough current to be concerned in the least. I had used Schottky stud mounted, I seem to think 30 or 50 amp devices in the original version, but went with the HEXFREDs since everyone was crowing about how great they were. Which I suppose they are. Given the huge charging currents, I thought they would be a potential benefit.
(could not detect any audible difference)
_-_-
DF96, I remain suspicious of the little SMD inductors... would still like to see the current waveform under dynamic conditions across these devices, in the actual PS.
It seems that the goal of this PS design is to not let the impedance rise as frequency rises. Then the added C at the end is to provide an additional "reservoir" given that any HF stuff has been "cleaned" off first. This "local filtering" idea has been touted by some commercial manufacturers who put caps right next to the output devices. All sorts of good, interesting reasons have been given and all sorts of more or less BS reasons have been given.
I always thought that with the demise of large filter caps (for the most part) that manufacturers had to come up with a marketing ploy to justify the use of what was available - lots of smaller caps in parallel.
I think the idea, in concept is that with a relatively high impedance filter being the source, you add some "stiffness" at the end of the string by the lump amount of C that hangs before the output devices. Clearly, only useful within the time constant required for peak draw and recover...
Again, I'd like to see some of the measurements of the actual amp in operation that I had suggested.
The PSUD2 simulation is also a good idea.
_-_-
It seems that the goal of this PS design is to not let the impedance rise as frequency rises. Then the added C at the end is to provide an additional "reservoir" given that any HF stuff has been "cleaned" off first. This "local filtering" idea has been touted by some commercial manufacturers who put caps right next to the output devices. All sorts of good, interesting reasons have been given and all sorts of more or less BS reasons have been given.
I always thought that with the demise of large filter caps (for the most part) that manufacturers had to come up with a marketing ploy to justify the use of what was available - lots of smaller caps in parallel.
I think the idea, in concept is that with a relatively high impedance filter being the source, you add some "stiffness" at the end of the string by the lump amount of C that hangs before the output devices. Clearly, only useful within the time constant required for peak draw and recover...
Again, I'd like to see some of the measurements of the actual amp in operation that I had suggested.
The PSUD2 simulation is also a good idea.
_-_-
Yes. He appears to be worrying about the (alleged) PSU behaviour at 100MHz, while remaining blissfully unaware that behaviour at 100mHz is more important: any 100MHz weaknesses can be fixed by local decoupling, but to fix 100mHz you would need a regulator - and that would bring in a new set of issues to think about.bear said:
I am not sure how. Can you tell me more on how to connect the scope probe and ground lead?
Perhaps I could use probe1 to measure one side of the inductor and probe2 the other side, and both ground leads connected to the PSU ground, then use some math function to display the differences between the two?
Last edited:
Thank you very much, Bear and DF. Your advice has been extremely helpful and will continue to be very much appreciated.
I think next I will experiment some changes to the PSU and to do some measurements, hopefully during the coming weekend.
I think next I will experiment some changes to the PSU and to do some measurements, hopefully during the coming weekend.
One method to try is to use the "add" function between two channels.
The other method that while I personally use, is considered UNSAFE is to float the scope with a small isolation transformer and then you can put the ground lead anywhere. So, I can not recommend this. Use at your own risk. 😀
The best method is a scope with a differential probe, but since neither you nor I have one, that's a no-go.
But you don't NEED this if you are putting the scope ground ON GROUND and measuring away from that ground point along the bus - especially if there is some current being drawn, you'll see something show up.
_-_-bear
The other method that while I personally use, is considered UNSAFE is to float the scope with a small isolation transformer and then you can put the ground lead anywhere. So, I can not recommend this. Use at your own risk. 😀
The best method is a scope with a differential probe, but since neither you nor I have one, that's a no-go.
But you don't NEED this if you are putting the scope ground ON GROUND and measuring away from that ground point along the bus - especially if there is some current being drawn, you'll see something show up.
_-_-bear
A capacitor resonating with capacitors? Wouldn't you need an inductance? And where is that highQ cap in the schematic?
Why do you think there is resonance? Did you actually measure impedance?
have you ever considered floating the device under test? worthwhile and done by 99% of the power labs everywhere.
I disconnect the earth wire of the DUT during tests, and my scope is not connected to earth but powered by battery. Otherwise I found more mysterious RF noise.
RF ? still ??
You did not post the amp's schema ... but you have two pole comp and a
huge amount of lead compensation (22pF). You might actually be oscillating ?
I see the cascodes . If I simulate my leach (spooky) with that 22p , a huge
resonate spike at @ 1mhz appears.
Just pull the 2 resistors on the TPC and eliminate the 22p.
Then you will a "plain jane" miller compensated leach amp as a baseline.
OS
You did not post the amp's schema ... but you have two pole comp and a
huge amount of lead compensation (22pF). You might actually be oscillating ?
I see the cascodes . If I simulate my leach (spooky) with that 22p , a huge
resonate spike at @ 1mhz appears.
Just pull the 2 resistors on the TPC and eliminate the 22p.
Then you will a "plain jane" miller compensated leach amp as a baseline.
OS
Yes. But it takes a larger iso transformer? 😉
And, you have to also watch out for the input source grounding the chassis, even though the DUT is floating!
_-_-bear
Screen shot of the scope?
Have you looked at your baseline RF pickup?? (for starters at the same probe locations and the amp off??)
Generally speaking anything under say 7mv p-p of noise on a scope looking at a power amp out is "pretty good" and what's there is hard to discern between random noises, RF noises, radiated "hash" and who-knows-what.
Putting a LP and then a HP filter on the input of the scope probe will get rid of part of the spectrum and maybe give a more meaningful look. Maybe.
What's the amplitude of this mysterious signal??
_-_-bear
I asked your thoughts. Do you know what a resonance is, and how to calculate it? Did you actually measure it (instead of electromagnetic interference from anywhere)? Did you try to embed the neccessary elements into your simulation?
I didn't see anything in this thread suggesting any of the answers is yes.
RF pickup is not mysterious; it is ubiquitous. Unless you are in a screened room then whenever you look for RF you will find it.HiFiNutNut said:
- Status
- Not open for further replies.