Well Keit - you are at liberty to paint yourself into a corner by refusing to use any solid state devices in your amplifier designs (but - have we seen any amplifier designs contributed by you?) - But the argument over whether transistor current sources, cascodes, FET grid-drivers, DHT filament regulators, preamp shunt-regulators, and many other solid-state assistance circuits can improve valve amplifiers (by measurement, or listening) has been won years ago. If you are in any doubt about the opportunities an open-minded approach to device-selection can yield, just look up any of the subjects I have named.
Clearly you are also confused about pulse shapes and spectrum. If not, you would not think that rectifiers, whose reverse-recovery time ranges from tens to hundreds of nanoseconds - could confine their spectrum to "way lower than 10MHz". But I am probably misreading you because you are confining your definition of rectifiers to those produced before 1949.
Clearly you are also confused about pulse shapes and spectrum. If not, you would not think that rectifiers, whose reverse-recovery time ranges from tens to hundreds of nanoseconds - could confine their spectrum to "way lower than 10MHz". But I am probably misreading you because you are confining your definition of rectifiers to those produced before 1949.
Nobody is trying to misrepresent you, Keit. But you have come along and made nit-picking arguments based on narrowly-defined criteria of your own imagination, and been indiscreet and offensive in the process.
Folks here are not misrepresenting you, we are saying
'Make constructive, polite and useful contributions - or shut up'
Folks here are not misrepresenting you, we are saying
'Make constructive, polite and useful contributions - or shut up'
Well, a pulse lasting 10's to hundreds of nanoseconds is of the megahertz region, not the 100 Mhz region you squawked abaout. But you missed my points:-
1) If its megahertz you can't hear it (yes I know you claimed it can have audible effects - your just trying to defend the undefendable
2) Some thing that occupies 10's or hundreds of nanoseconds each 10 milliseconds (a ratio of >100,000:1 or 100dB) cannot put out much power to cause a disturbance.
The recovery time trr of a 1N4004, quite fast for a power rectifier, is 2 microseconds, not 10's of nanoseconds, and it is still a current type.
Keit you contention that substituting a simple resistor of equal value to the DCR of a ballast was empirically shown to be false - yet you repeat the assertion that a resistor can perform as well. That shows a lack of ability to assimilate empirical evidence which about sums up what I am reading from you.
Shoog
Shoog
Yes, you make a proper measurement of incremental inductance whereas my method yields something in between amplitude and incremental inductances (but skewed towards incremental).
In order to make a "proper" measurement, I would need to add an additional source of modulation, but my setup wasn't built like that.
It was mainly intended for ferrite inductors for which the difference is very small (because of their low hysteresis compared to metal iron).
In addition, when the core is gapped, the hysteresis error becomes "diluted", and most power chokes have some form of gapping, sometimes very substantial (and anyway, for this kind of component, you're not looking for 1% accuracy)
Of course a choke performs better than a resistor, the point is that in this age of cheap large value electrolytic capacitors, you don't really need the extra performance of the choke.
Just because a rectifier's reverse recovery transient lasts several microseconds, doesn't mean that it contains no high frequencies. It tends to end very abruptly unless the diode has been deliberately processed for soft recovery, and this sharp cutoff can generate as high frequencies as you like.
Just because a rectifier's reverse recovery transient lasts several microseconds, doesn't mean that it contains no high frequencies. It tends to end very abruptly unless the diode has been deliberately processed for soft recovery, and this sharp cutoff can generate as high frequencies as you like.
Resorting to large cheap electro's means that the charging spike and the recovery spike are generally of a much higher magnitude than with smaller caps. This can get to the stage where the transformer dumps the energy as a radio pulse which is all but impossible to filter out since it pollutes everything. For this reason alone I would always choose to build a CLC power supply with the first cap been less than 10uf.
Big caps have a place after the choke where they provide limitless power delivery potential. A fluro ballast is enough to make a valve rectified power supply with 4uf > ballast > +400uf of capacitance - immune from arcing issues. This should tell you everything you need to know that it is doing more than a 33R resistor.
Big caps are not a substitute for a choke.
Shoog
Big caps have a place after the choke where they provide limitless power delivery potential. A fluro ballast is enough to make a valve rectified power supply with 4uf > ballast > +400uf of capacitance - immune from arcing issues. This should tell you everything you need to know that it is doing more than a 33R resistor.
Big caps are not a substitute for a choke.
Shoog
Last edited:
Exactly, Steve.
The attached picture shows (blue trace) a diode with ~ 250ns recovery time, but an obvious spectral component at ~20MHz.
The actual spectrum will depend on the parasitic inductance and capacitances involved.
Attachments
The charging and recovery spikes are not that hard to filter out. Solid-state amp builders have been dealing with spikes an order of magnitude bigger for 40 years, while achieving lower noise and distortion than valve amps.
I still somewhat prefer Shoog's type of CLC filter in a valve amp, but for slightly different reasons. The spiky current due to a large first capacitor results in a poor power factor. By making the first cap smaller, the transformer will run cooler for a given DC output current.
I still somewhat prefer Shoog's type of CLC filter in a valve amp, but for slightly different reasons. The spiky current due to a large first capacitor results in a poor power factor. By making the first cap smaller, the transformer will run cooler for a given DC output current.
I have had quite the opposite experience with SS, I have found them considerably harder to filter than any of my valve amps which tend to have a much lower noise floor for exactly the reasons described.
Shoog
Well, I have seen one or two solid-state amps built with CLC power supplies. 🙂 Personally I have no problem getting a noise floor 100db+ below full output in a solid-state amp with the ordinary bridge rectifier and 2 capacitor power supply. The last valve amp I built only managed about 95dB due to traces of hum.
There you go again - misrepresenting what I said. I never said anything about the resistor being the same as the choke DCR, and everything else the same. That would be silly - if for the very least the better the choke the lower the DCR.
One would choose a resistor to do the job. You will note I wrote (more than once) words like"...especially now that large value electrolytics are available. In other words, rather than put a great ugly fluro ballast in, try a resistor with, if necessary to retain sufficiently low power supply droop, a larger capacitor.
As I said before, if it is desired to have a small electro, that will be valid if you are restoring an old amp or re-creating the look of an old amp. In both cases, using a fluro ballst makes no sense to me, not on electrical grounds but on appearance grounds. One does not restore a vintage car by putting and a modern alternator in.
Well, I have to admit you aren't picking on me - you misrepresented ScopeBoy too.
He didn't say filtering was easier in SS - he said the issue was worse than for tube amps, but SS designers never the less solved it easy enough.
Your first point - my point exactly!
Your second point - Yes, but the mathematical technique of Fourier Analysis shows us that the energy in the high frequencies generated by the abupt turn off must be low.
Last edited:
You are both spot on in all this.
There is another reason for small caps immediately following the rectifier if the rectifier is a vacuum tube. Tube rectifier ratings forbid the use of capacitors larger than the small sizes used in the pre-transistor era. Use larger caps and the tube can detructively arc over internally.
To prevent rectifier arc over, put most of the capacitance in the second electro, as usually done, and isolate it from the rectifier with either a choke or a resistor. Either will protect the rectifier. A starting value for the resistor is the lowest permitted transformer resistance given in the tube rectifier datasheet. It will be larger than 33 ohm, but quite often it's low enough not to affect load droop too much. Not all datasheets give this resistance, but you can interpolate from similar tubes for which it is given.
Last edited:
If you design a top plate to incorporate a Fluro ballast then it can look fine, and if you don't like it there put it under the lid. These are really petty quibbles.
Using a resistor to do the work of a fluro ballasts will add considerably to the B+ requirements and waste more energy to boot - they are not equivalent in any meaningful way. You have just admitted that a fluro ballast brings more to the power supply than its simple DCR equivalence which is the whole point of this discussion in the first place.
This is really coming down to a personal aesthetic bias against fluro ballasts.
Shoog
Using a resistor to do the work of a fluro ballasts will add considerably to the B+ requirements and waste more energy to boot - they are not equivalent in any meaningful way. You have just admitted that a fluro ballast brings more to the power supply than its simple DCR equivalence which is the whole point of this discussion in the first place.
This is really coming down to a personal aesthetic bias against fluro ballasts.
Shoog
Last edited:
Fourier analysis says nothing of the sort, it simply says that the more abrupt the turnoff, the more high frequency energy will be generated, and the higher the frequency content of it. There are purpose-built step recovery diodes that can generate energy up to 10s of GHz.
On the original topic, one thing to ponder is that salvaged fluoro ballasts may well be partly cooked and full of shorted turns.
Here's a question, where you comparing like with like. Was your valve amp class A and your SS amp class AB.
My experience with comparing Valve with SS is like for like with class A designs. It can be brutally difficult to get a quiet supply when delivering 2 amps of continuous draw from a SS power supply without recourse to chokes. Charging transients become a serious issue as they tend to bleed into the ground plain where they get everywhere. The size and lengths to which you have to go to get a SS Class A power supply’s can be frankly ridiculous.
Its relatively simple to get better performance for the same energy output with a CLC valve amp. The lower current's are the obvious reason.
Shoog
The valve amp was a very hot biased class AB and the SS one was a regular class AB.
I still think it is comparing like with like, as solid-state doesn't need Class-A to achieve better performance than any valve amp.
Also I showed with a distortion analyser that the SS amp was still successfully rejecting interference from the power supply while putting out its rated power. This is of course thanks to the evils/wonders of global feedback. 🙂
Charging transients may well be visible when probing grounds with a scope, but the real issue is whether they show up in the residual output of the distortion analyser.
I still think it is comparing like with like, as solid-state doesn't need Class-A to achieve better performance than any valve amp.
Also I showed with a distortion analyser that the SS amp was still successfully rejecting interference from the power supply while putting out its rated power. This is of course thanks to the evils/wonders of global feedback. 🙂
Charging transients may well be visible when probing grounds with a scope, but the real issue is whether they show up in the residual output of the distortion analyser.