Sad fact of life: good stuff cost $$ - $$$. In your case, I see 2 major problems:
1.- Instrumentation: a vanilla 44KHz sound card won't cut it, as mentioned above. A $100 Focusrite Solo is good to about 90KHz (192KHz/2). Better still E1DA's COSMOS Scaler + ADCiso will let you reach 180KHz, for more $. My preferred software is REW (free). At least you'll be able to measure the 1st stage to make sure it's not its fault.
2.- Output transformer: this is the most difficult part, and most sellers don't publish detailed data. As a rule of thumb, a mid-fi push-pull OPT will have resonances around 50-100KHz and cost about $500 for a pair. A single ended OPT is a lot harder to make because of the mandatory air gap, so expect to pay more for less. If you're sure your current OPT is the culprit, and since you're in Poland, I'd suggest you check out toroidy.
1.- Instrumentation: a vanilla 44KHz sound card won't cut it, as mentioned above. A $100 Focusrite Solo is good to about 90KHz (192KHz/2). Better still E1DA's COSMOS Scaler + ADCiso will let you reach 180KHz, for more $. My preferred software is REW (free). At least you'll be able to measure the 1st stage to make sure it's not its fault.
2.- Output transformer: this is the most difficult part, and most sellers don't publish detailed data. As a rule of thumb, a mid-fi push-pull OPT will have resonances around 50-100KHz and cost about $500 for a pair. A single ended OPT is a lot harder to make because of the mandatory air gap, so expect to pay more for less. If you're sure your current OPT is the culprit, and since you're in Poland, I'd suggest you check out toroidy.
OK, you should be able to use a cliplead attached at the coupling cap out or the grid resistor, brought out from under the amp and connect that to the cable to the soundcard. It should work the same using the same ground point, too.
And as an aside, I think your FR is doing as well as can be expected with that amp. It is actually biased to be 1/2 of an AB push-pull but using a single-ended OPT. Not using a conventional class A operating point.
And as an aside, I think your FR is doing as well as can be expected with that amp. It is actually biased to be 1/2 of an AB push-pull but using a single-ended OPT. Not using a conventional class A operating point.
Unfortunately, in order to test the frequency response of the input circuit and driver tube, the grid of the KT88 is in the middle of the KT88 plate to driver tube plate negative feedback loop. So . . .
Disconnect the feedback resistor from the KT88 plate, and connect it to the same point as the driver tubes plate load resistor (now 2 resistors in parallel).
Connect a proper load resistor across the amplifier common and 8 Ohm tap.
Connect your signal generators signal to the amplifier input; first 1 kHz, then 20kHz.
Use a low level generator signal, the amplifier's total gain went up because now there is no plate to plate negative feedback.
Be sure you send exactly the same generator volts at 1kHz as you do at 20kHz.
Connect your (properly frequency compensated) scope probe to the grid of the KT88.
Measure the KT88 grid volts at 1k and 20k.
20Log (20kHz volts / 1kHz volts) = dB rolloff at 20kHz.
That is the frequency rolloff of the amplifier input circuit, driver stage, and the KT88 grid capacitance.
The KT88 grid capacitance is greater in Triode Wired mode, than it is in Ultra Linear Mode.
Miller effect capacitance is greater when the screen moves as far [100%] as the plate [Triode wired mode].
Miller effect capacitance is less when the screen moves only 40% as far as the plate [Ultra Linear mode].
By the way, I never use a 100k potentiometer volume control; I use 50k.
For high gain input/driver triodes, the miller effect capacitance might cause a slight high frequency roll off at some settings of the volume control, and with some signal source output impedances.
Of course, your signal source has to be capable to properly drive 50k potentiometer, versus the ease of driving a 100k potentiometer.
Disconnect the feedback resistor from the KT88 plate, and connect it to the same point as the driver tubes plate load resistor (now 2 resistors in parallel).
Connect a proper load resistor across the amplifier common and 8 Ohm tap.
Connect your signal generators signal to the amplifier input; first 1 kHz, then 20kHz.
Use a low level generator signal, the amplifier's total gain went up because now there is no plate to plate negative feedback.
Be sure you send exactly the same generator volts at 1kHz as you do at 20kHz.
Connect your (properly frequency compensated) scope probe to the grid of the KT88.
Measure the KT88 grid volts at 1k and 20k.
20Log (20kHz volts / 1kHz volts) = dB rolloff at 20kHz.
That is the frequency rolloff of the amplifier input circuit, driver stage, and the KT88 grid capacitance.
The KT88 grid capacitance is greater in Triode Wired mode, than it is in Ultra Linear Mode.
Miller effect capacitance is greater when the screen moves as far [100%] as the plate [Triode wired mode].
Miller effect capacitance is less when the screen moves only 40% as far as the plate [Ultra Linear mode].
By the way, I never use a 100k potentiometer volume control; I use 50k.
For high gain input/driver triodes, the miller effect capacitance might cause a slight high frequency roll off at some settings of the volume control, and with some signal source output impedances.
Of course, your signal source has to be capable to properly drive 50k potentiometer, versus the ease of driving a 100k potentiometer.
Last edited:
checked FR at the KT88 grid and it looks ok... I've also noticed that the KT88 plate voltage is a bit lower than in the desing - around 425V (not sure if it matters here)
so what do I make out of it? thx
so what do I make out of it? thx
That looks great. So I would not expect the rolloff to be tube related in the audio range. That only leaves the OPT, which if it is 5K primary would be about 2x what a KT88/6550 used in a single ended output would expect. I don't know if that may create the rolloff but the OPT seems to be where it originates and without using GNFB the amp is going to be limited by OPT issues. But I really don't think you'll have a sound quality issue at all, just power. Good luck with your project. This is an experimenter's playground amp. Enjoy.
@jank9
Read the posts here. Posts 3516-3519. The moral is this. It really doesn’t matter who the transformer manufacturer is or how much money you have paid or what guarantees they give you on their websites. You need to test PRIOR to installation. I had to return both my Edcor CXSE25-5K transformers as one was clearly out of spec. Yours appear to be quite good in all honesty. They replaced my pair without fuss, but I had to show them the measurements (100Hz increments starting at 5khz).
Best,
Anand.
Read the posts here. Posts 3516-3519. The moral is this. It really doesn’t matter who the transformer manufacturer is or how much money you have paid or what guarantees they give you on their websites. You need to test PRIOR to installation. I had to return both my Edcor CXSE25-5K transformers as one was clearly out of spec. Yours appear to be quite good in all honesty. They replaced my pair without fuss, but I had to show them the measurements (100Hz increments starting at 5khz).
Best,
Anand.
No idea.
It's just an example to show how bad it can be, while staying within the +/-3dB limit, 10-45000Hz in this case.
I would just enjoy the amp. You aren't gonna hear 1/2db at 10K and 1db at 15K.
I am with Stephe. Even 3dB is barely noticable to most folks especially at such extreme high frequencies.
Also with Stephe. I built the Kegger/Blueglow using the recommended Edcor CXE25-8-5K OPTs. I run a 12AT7 for the input and use a feedback resistor increased to 280k ohms.
I have Polk SDA speakers and they are a little bright at the top so maybe it is a good match. Sounds really good to me.
I have Polk SDA speakers and they are a little bright at the top so maybe it is a good match. Sounds really good to me.
Last edited:
Guys, thanks a lot again for very interesting comments.
@Zung, @poseidonsvoice - thanks for the reference - something I was hoping for to help me put the OTs I have in some perspective.
@Zung, @poseidonsvoice - thanks for the reference - something I was hoping for to help me put the OTs I have in some perspective.
If I did however want to upgrade the OTs (i know its a rabbit hole and i should enjoy listening, but with DIY tweaking is as much fun as listening) what experience do you guys have with Lundahl? They have a LL2769 model appearntly designed for the likes of KT88s, but it has 4,7k primary impedance (while amp design asks for 5k). Is 4,7k OK? What could be the conseuence of lower primary impedance?
4.7K or 5.4K, not enough difference to matter. I'd be interested to see if the 15W 5K Edcors do this..
@jank9
I have listened to Lundahl transformers on numerous occasions. Amorphous Core or Nanocrystalline, are both excellent and should be a nice upgrade over Edcor. Even standard version is better imho than Edcor.
That being said, the LL2769 is appropriate for SE designs up to 20W (it’s used in the Elekit line of kits). You may want to look through Lundahl’s catalog and get a transformer that is more appropriately sized for your KT88 SE since some of the smaller output transformers are also available in amophous core and nanocrystalline core.
The nice thing about Lundahl is quality control, repeatability, reliability and a very flat frequency response. Every one I have measured has been excellent. Their process is quite rigorous. But you pay for it. My other go to transformer is Monolith Magnetics. A measurement driven company which I like. Again, expensive. So I use it only in the best fully vetted tube circuit designs. Not prototypes!
Finally, you might consider Toroidy which is local to you in Poland. I use their power transformers for my solid state builds quite a bit, but I’ve heard positive reports on their output transformers too. There are fans of toroid based output transformers, you may want to ask them.
Best,
Anand.
I have listened to Lundahl transformers on numerous occasions. Amorphous Core or Nanocrystalline, are both excellent and should be a nice upgrade over Edcor. Even standard version is better imho than Edcor.
That being said, the LL2769 is appropriate for SE designs up to 20W (it’s used in the Elekit line of kits). You may want to look through Lundahl’s catalog and get a transformer that is more appropriately sized for your KT88 SE since some of the smaller output transformers are also available in amophous core and nanocrystalline core.
The nice thing about Lundahl is quality control, repeatability, reliability and a very flat frequency response. Every one I have measured has been excellent. Their process is quite rigorous. But you pay for it. My other go to transformer is Monolith Magnetics. A measurement driven company which I like. Again, expensive. So I use it only in the best fully vetted tube circuit designs. Not prototypes!
Finally, you might consider Toroidy which is local to you in Poland. I use their power transformers for my solid state builds quite a bit, but I’ve heard positive reports on their output transformers too. There are fans of toroid based output transformers, you may want to ask them.
Best,
Anand.
Last edited: