Kegger/blueglow kt88 project

[TriodeLuvr]

I have been using the same nos mullard 5ar4 for over 2.5 years. My second power supply cap is nichicon 390uf 500v. I also did not think the large cap mattered after the choke?

I don’t want to offend anyone’s sensibilities, but I would like to review a few power supply basics….

The first purpose of the CLC filter is to provide good attenuation at the ripple frequency (100 Hz or 120 Hz). However, the surge in current that occurs when the 5AR4 first warms up represents a much longer time constant (lower frequency) than that. The current will flow into the filter caps in direct proportion to the impedance represented by the filter section at that lower frequency, and this impedance in turn depends on the values and LC ratio. Small parallel C and large series L will reduce the current, and vice versa. In the case of the amplifier under discussion, the inductance of the choke does little to reduce the surge into the last cap.

The second purpose of the CLC is to present a low impedance to the power amplifier (looking back into the power supply) so that the power amplifier will be supplied a constant DC voltage. This low impedance must be maintained by the power supply over the entire operational frequency range of the amplifier, otherwise, the B+ of the amplifier will be modulated by the frequencies it's amplifying. AB1, AB2 and A2 amplifiers also attempt to modulate the power supply at infrasonic frequencies, because they draw more or less current from the supply over the longer time periods represented by the transition from one power level to another. These classes draw more current when they are driven harder, and these ultra-low frequency current variations must also be met with a low power supply output impedance, otherwise the output voltage changes. The simplest means to managing these low frequency demands is to increase the size of the second CLC capacitor. Because it responds directly to the low frequency current demands, the last cap is often referred to as a reservoir capacitor. The larger it is, the longer it can maintain a low impedance and support extended periods of high output power.

Unlike the other amplifier classes, current drawn by a SET/SEP operating in class A1 changes very little from zero output to full power. A large reservoir capacitor isn't needed, because we only need to maintain a low power supply impedance across (approximately) the audio spectrum we are amplifying. For this reason, a much smaller capacitor can be used at the output of the supply. There's no basis for assigning typical reservoir cap values here, and certainly no reason to apply a filter design that stresses the rectifier.

OK, those are the basics. To be more specific, I used PSUDII to look at the current flow in this power supply. The surge through the 5AR4 is way over its specification with the 200uF cap in place. I also used SPICE earlier today to plot the power supply's output impedance over frequency. The smaller capacitor I suggested earlier is absolutely sufficient for good performance, and I would urge the OP to try this first. If a known-good rectifier still fails, something additional is indeed wrong. However, changing the cap will at least eliminate this issue as the reason for the failures and guarantee the reliability of your amp once everything else is squared away.

Jack

 

[Simon Murray]

Jack - thanks for that. No sensibilities offended. But it does put the cat amongst the PS pigeons a little!

I guess you know there's a bit of history and development in this amp. Alex Gendrano's 2007 version had 10uF before the choke and 100uF after. Kegger recommended a Solen 10uF before and between 180 and 470uF after (Kegger sometimes used two 470uF caps in series "for roughly 235uf, [because] 200uf is kinda an amount I shoot for as a "decent" size in that spot"). I saw that MWHouston (in an EL34 version) used 25uF before and 2 x 47uF after. Mark (Blueglow) who summed up the history and did a nice video build series (which I followed) used 2 x 100uF and specifically comments (see video 11, around the 50 minute mark) on the amount of capacitance that rectifier tubes can deal with in the first stage of the CLC filter circuit, and commented relevantly that "by the way, this tube doesn't care about the capacitance on the other side of the inductor here, just the first one."

I tried to do my own research on the point. The only reference I could find in 5AR4 spec sheets was to the maximum capacitance at the input of the filter circuit: eg see here: "Condenser Input to Filter (Max) .... 60uF" or here: "Maximum Capacity (condenser input filter) .....60uF" or here: "Capacitor Input: C = 60uF" and here: "[for both Maximum Design Value and for Typical Operation] Input Capacitor of Smoothing Filter: 60uF [for all specified voltage ranges]"; and plenty more.

I don't have the expertise to comment on this myself but have to follow the accepted wisdom. So I'm presently using the two parallel 100uF caps as per the Abdellah/Gendrano/Kegger/Blueglow circuit. Am I likely to blow my 5AR4?

I'm sure there are quite a lot of us who have used this circuit and are in the same position.

I don't want to seem to be pushing back (because i don't have any real experience or knowledge myself) but can you point us to a spec sheet or a text that confirms your position, and would prompt us to modify our circuits? That would be most instructive.

Simon

 

[TriodeLuvr]

Hi Simon, thanks for your comments. The process of comparing a specific combination of component values to the ratings isn’t rocket science, but it’s time consuming. Take a look at Rating Chart I, II and III in the GE datasheet to see what I mean. Anyone who wants to do this manually has my respect.

https://frank.pocnet.net/sheets/093/5/5AR4.pdf

A simpler means of accomplishing this is the use of the simulator I mentioned earlier, PSUDII (Power Supply Designer II), available free at the Duncan Amplification site. This program contains accurate models of several tube rectifiers, including the 5AR4. After you enter all the component values, it will alert if the rectifier’s ratings have been exceeded. Then you can view voltage, current etc. in various parts of the circuit to see for yourself what's happening.

There are numerous variables in all this, including primary and secondary DCR of the power transformer, resistance of the choke, ESR of the capacitors, line voltage etc. Entering data into PSUD might not trap all these characteristics, but it will get close. If the program doesn’t alert to excessive current or voltage immediately, it’s a simple matter to change one of the factors - say the transformer resistance - to see how close it is. Or, view the plots for various nodes in the circuit and compare what you see to the ratings charts. And on this point, the program doesn’t know about the vagaries of all the various reissue tubes. It’s necessary to use it with caution in this regard and to be sure a large margin of safety is included in selecting final component values.

I’ve been designing choke- and capacitor-input supplies for quite a few years. If this circuit had a final cap value under 100uF, or if the choke was 20H with typical DCR, I wouldn’t have even poked my nose in. But the fact is, I can just look at the values here and see that they’re on the wrong side of what a 5AR4 should be expected to do. Whether the OP has a failed component, or his environment and specific combination of components is too close to the edge, I can’t say for sure. However, I highly suspect the latter, and that’s why I chimed in.

As a final comment, someone earlier suggested adding series resistance to reduce the peak current demands on the 5AR4. That’s not a bad idea in a class A1 SET, because PS regulation isn’t such a significant factor. So, maybe that's another topic for another day.

Jack

 

[Simon Murray]

Thanks again Jack and 6DZ7 for this alert and comments. Would you mind giving one more suggestion?

I want to reduce the risk to my 5AR4 I've inadvertently created. My amp uses a F&T dual 100uF cap run parallel. An easy option is disconnect one of the caps - leaving 100uF. Do you think that's enough of a reduction or should I go further? What will I notice if I reduced the second cap too much - 50 cycle hum due to inadequate filtering and/or lost power or distortion on the peaks?

Don't think it matters but I'm actually using Hammond iron equivalent to the Edcor specified.

Thanks again for your thoughts ..

Simon

 

[TriodeLuvr]

Thanks again Jack and 6DZ7 for this alert and comments. Would you mind giving one more suggestion?

I want to reduce the risk to my 5AR4 I've inadvertently created. My amp uses a F&T dual 100uF cap run parallel. An easy option is disconnect one of the caps - leaving 100uF. Do you think that's enough of a reduction or should I go further? What will I notice if I reduced the second cap too much - 50 cycle hum due to inadequate filtering and/or lost power or distortion on the peaks?

Don't think it matters but I'm actually using Hammond iron equivalent to the Edcor specified.

Thanks again for your thoughts ..

Simon
View attachment 2311961

Reducing the value to 100uF will definitely reduce the surge through the 5AR4. There should be no issues. The filter will remain very effective at attenuating ripple, and bypassing at the second cap will still be more than adequate across the audio spectrum. Don't confuse this with a push-pull amplifier that draws more current at higher output (although this is enough capacitance for that, as well). The only requirement here is audio frequency decoupling.

Jack

 

[Simon Murray]

Thanks 6DZ7 - yes indeed, Hammond 374BX 750v CT power tranny - along with a 193J 10H 200mA choke (and 1628SEA OPTs). Hammonds were mucho cheaper to ship to Australia than Edcor as specified in the BOM. My first cap before the choke is a 10uF 630v Solen as per the circuit.
Simon

And thanks very much, Jack, for your reply. Excellent. Might there some way I can readily test for surge at turn-on, before and after isolating half my 100+100uF cap. This issue will be of interest to quite a few other builders, I'm sure. I have a Fluke and a 'scope.
Simon

 

[Hyakuya]

Hi.

When I built this amp a year or so ago I had problems with the rectifier arcing and fuses blowing.

After many hours of troubleshooting the solution for me was to put a 10R resistor after the rectifier. I think it was a 10W surface mount but anyway it stopped the problem I had.

The essence being it didn't drop the voltage but did prevent the surge at startup. Easier than changing the caps so maybe worth a go.

Of course there may be reasons not to do this but I am in the keen amateur camp so happy to know otherwise....

Hope you get it sorted out though. It's horrible going through all the time and effort to build this and not have it working when you finish. It is worth the time to fix though. I love my amp...

 

[TriodeLuvr]

Might there some way I can readily test for surge at turn-on, before and after isolating half my 100+100uF cap. This issue will be of interest to quite a few other builders, I'm sure. I have a Fluke and a 'scope.
Simon

I use simulators exclusively for this sort of thing, but you could probably measure the surge if you have a digital storage scope. Lift the CT of the HV secondary and temporarily install a 0.5 ohm (or smaller) resistor in series from the CT to ground. Attach the scope probe across the resistor to capture the voltage spikes that occur when the amplifier is turned on. Then use Ohm's Law to convert that to current.

Jack

 

[TriodeLuvr]

Hi.

When I built this amp a year or so ago I had problems with the rectifier arcing and fuses blowing.

After many hours of troubleshooting the solution for me was to put a 10R resistor after the rectifier. I think it was a 10W surface mount but anyway it stopped the problem I had.

The essence being it didn't drop the voltage but did prevent the surge at startup. Easier than changing the caps so maybe worth a go.

Of course there may be reasons not to do this but I am in the keen amateur camp so happy to know otherwise....

Equipment built in the '50s and '60s frequently used a surge resistor directly after the rectifier, even when it wasn't a 5AR4. The value of the resistor was typically between 47 ohms and 100 ohms. Below is the power supply in a Stromberg-Carlson that powered a stereo amplifier with four 6V6s. There's absolutely nothing wrong with this approach.

Jack

 

[Simon Murray]

OK, cool. I looked at the specs for it. You are using both of the 120v primary windings in parallel as specified by Hammond for it, yes?

Yes - just so:

 

[Simon Murray]

.... but you could probably measure the surge if you have a digital storage scope. Lift the CT of the HV secondary and temporarily install a 0.5 ohm (or smaller) resistor in series from the CT to ground. Attach the scope probe across the resistor to capture the voltage spikes that occur when the amplifier is turned on. Then use Ohm's Law to convert that to current.

Jack

Very good, thank you Jack. I'll try that.
Simon

 

[Simon Murray]

Hi.
......
Hope you get it sorted out though. It's horrible going through all the time and effort to build this and not have it working when you finish. It is worth the time to fix though. I love my amp...

Hyakuya - thank you! To be clear, I am not presently having problems with blowing fuses or rectifiers. No problem at all, in fact. But Kentuckycat was having big problemos. And I wish to avoid having problems. So the hints and tips that have been shared are extremely useful for those of us without a lot of experience or built-up knowledge.

But I thought I should show my amp ...
Simon

 

[Simon Murray]

OK, it's 240v primary then. So if you run the factory internal resistance values for the primary windings and one winding of the secondary you get ~70R for the PT internal resistance, per plate. That's not sufficient for the output voltage and the input capacitance of the PS, per old school 5AR4's. If it was my amp, I'd first lower that second cap back down to ~47-50R. You could do that by wiring the double cap in series. But I'd also seriously consider adding 50R 5W resistors into the HT leads before the rectifier.

Thanks 6DZ7 - are you suggesting lower the cap and add resistors - or just do one or other (and you favour the resistor approach)? If I can work how to get my scope to check the surge I can see the effect of this mod, too.
Simon

 

[Simon Murray]

shiny fishbait mods

Hahaha - fell off my chair!

Incidentally, maybe I do have one tiny niggle that someone may be able to comment on.

I think the amp has a lovely sound with very little distortion. But at high volumes in U/L mode (but not so high as to distort/clip a 1K Hz sine wave into 8ohm dummy loads) I saw some ringing on one side only of a 1K Hz square wave. Pic below shows about 16 - 17v p-p output (8.5v p, or 6v rms, so close to 6 x 6 / 8 = 4.5W rms per channel). Is there an obvious cause for this bottom-only overshoot (it's not push-pull)?
Simon

 

[tubetime]

I don't believe most simulators take into consideration the slow startup of the 5ar4.

Every time you half the capacitance of the second cap you double the dc ripple going to the output tube. Not sure how much ripple you can have before you hear the hum.

 

[Simon Murray]

Well Jack, I lifted the CT and interposed 0.6 ohm resistor with two voltmeters across it. A Fluke to measure max volts and another to watch the voltage ramp up. Dummy loads on, no input.

Well, ramp up is not fast - like 15 seconds or more (I don't think this is slow metering). Anyway, the 1A slow-blo fuse crackled a bit then blew - I use1A slo for 240v, rather than 2A specified for 120v. The fluke showed a max 0.176v over 0.6R = 293mA. Anyway, nearly 300mA is quite a bit more than the 200mA spec for a Hammond 374BX and no doubt taxed the primary a little.

Ouch and weird ... how a little resistor causes this problem, when there's no problem with no resistor. Could to make you say "leave well-enough alone"? Maybe the 5AR4 doesn't like being run inverted, lol.

Anyways, replaced the fuse and slowly ramped up the AC on the variac - the amp's on speakers now. Looks normal, plays normal. Off.

Wait 5 mins - power up again, on the switch this time. Normal. So, off.

Wait 5 mins - power up again on the switch, this time watching the B+ rise. It's graceful ...takes about 20 secs to reach 300v and another 10 to reach 443v.

Doesn't seem like an avalanche to me. Thanks for your help.

Simon

PS: my set up:

 

[TriodeLuvr]

I don't believe most simulators take into consideration the slow startup of the 5ar4.

Every time you half the capacitance of the second cap you double the dc ripple going to the output tube. Not sure how much ripple you can have before you hear the hum.

According to PSUD, ripple with a 47uF final cap will be approximately 200mV pk-pk (120 Hz). Here's what that means to me:

The 5K:8 transformer will step that down to 8 mV. If we are using, say, 18 dB of feedback, this will be further reduced to 1.0 mV. Finally, converting pk-pk to RMS gives 0.35 mV. At 8 ohms, this produces 1.53E-08 watts, or a power level that is 78 dB below 1W. This means that if a woofer produces an SPL of 100 dB at 1W, this amount of ripple will produce 22 dB SPL at 120 Hz at one meter from the cone or horn mouth. This can be scaled to accommodate the actual sensitivity of the woofer. Also, it's late at night, so maybe someone should double-check my math.

PSUD has an optional soft start feature that ramps up the voltage over a period of 1/2 second. This produces numbers closer to real world results.

Jack

 

[TriodeLuvr]

Simon, this can't be measured with a voltmeter. The surge consists of a series of repetitive spikes that the voltmeter attempts to average. This produces a number well below the actual peak forward current. Here's what it looks like in PSUD:



It makes no sense that the fuse would blow. A small resistor in that position should not create any issues whatsoever. Something else is to blame for that.

Jack

 

[mfrischholz]

Simon, this can't be measured with a voltmeter. The surge consists of a series of repetitive spikes that the voltmeter attempts to average. This produces a number well below the actual peak forward current. Here's what it looks like in PSUD:

View attachment 2312860

It makes no sense that the fuse would blow. A small resistor in that position should not create any issues whatsoever. Something else is to blame for that.

Jack

Hello Jack,
trying to replicate your simulation with the exact same values you have used. For some reason I got different results than you, in particular I received no warning. Any idea why our results might differ? I am using PSUD II ver 2.20 build 74. Even used a clean install of PSUD to make sure all other settings are out-of-the-box.




Manfred

 

[tubetime]

Believe math is wrong, no global negative feedback.

Half second soft start is much different than real world 5ar4. Probably closer to 30 seconds like Simon measured.

Like I said, going on over 2.5 years playing daily 390uF for second cap and never an issue.