Can this be measured with a signal generator and scope? If so, what is the procedure?
Sure, first connect an 8R 50W dummy load to the 8 Ohm tap and common.
Warm up and bias the amplifier. Connect the (properly compensated)
scope probe to the 8R tap, and the ground clip to the common tap.
Input a 0.1V (approx) 1kHz square wave.
The standard circuit will have no overshoot at all. The tuning can be affected
by the age of the 7199 input tube. Never do this test near full output,
since most tube amps are not designed for that.
Last edited:
I appreciate all the replies fellas, but back to my original question, I'd like to extend the roll off of this stability network so that it doesn't adversely effect the upper end of the audio range, and obviously without harming stability. The original cap is 82pf, which after plugging it into the formula, gives a roll off frequency of about 9K. This is well within the audio range. I have calculated the poles of other standard capacitor values as shown below. Please advise. BTW, I reduced the 20db GNFB (of the original ST-70) to about 15 db, as this may be a factor.
Cap - pole F
82pf - 8826
75pf - 9650
68pf - 10644
62pf - 11674
56pf - 12924
51pf - 14192
47pf - 15400
43pf - 16832
Cap - pole F
82pf - 8826
75pf - 9650
68pf - 10644
62pf - 11674
56pf - 12924
51pf - 14192
47pf - 15400
43pf - 16832
Sounds like you want to change the feedback because you have the impression that it can't be as good as it could be due to the shelf network's low start corner frequency.
How are you going to determine if you've made an 'improvement'? Do you have the measurement and design skills to follow this through adequately, or is this just a bit of a learning curve effort (which is good in its own right)?
It seems like you're only doing a very simple design assessment, rather than having a go at assessing stage gains and phase plots and determining parasitic circuit values such as plate resistance, and next stage capacitance loading.
Do you have measurement equipment that can provide spectrum gain and phase, and harmonic and IM distortion? Does your existing amp meet the standard specs for output power, distortion and noise?
Just trying to give a heads up on the path you seem to be embarking on
How are you going to determine if you've made an 'improvement'? Do you have the measurement and design skills to follow this through adequately, or is this just a bit of a learning curve effort (which is good in its own right)?
It seems like you're only doing a very simple design assessment, rather than having a go at assessing stage gains and phase plots and determining parasitic circuit values such as plate resistance, and next stage capacitance loading.
Do you have measurement equipment that can provide spectrum gain and phase, and harmonic and IM distortion? Does your existing amp meet the standard specs for output power, distortion and noise?
Just trying to give a heads up on the path you seem to be embarking on
Do you even know what this circuit DOES?
Plot it out. The pentode with just a 200K load has a gain of say 200. If another 20K load is added, it has a gain nearer 20. The capacitor is open for low frequency and short for high frequency. Therefore we have gain of 200 flat from some very low freq up to some point, and gain of 20 flat above some other point. The two points ARE the two freqs you calculated.
Is that all? No. No amplifier is flat to infinite freq. That "gain of 20" must finally roll off to zero. This point is caused by "hidden" capacitances, tube and wiring strays.
But wait, there is more. The output transformer is a wicked maze of stray inductance and capacitance and resonance. The upper limits are greatly affected by the lower (bass) limit. 20Hz means BIG winding, >20KHz gets lost in there.
You gonna have to plot/measure ALL the reactances and gains, and compare to Nyquist Stability Criterion.
Look at your own numbers. Round to 10KHz and 100KHz. You have the high gain all the way through the audio band into the top octave. Removing the R-C network will only give 5dB more NFB at 20KHz. I'd say it is probably as good as it gets already, or so close we can't hear any real change. (Note that simple Harmonic Distortion on >10KHz is nominally "inaudible", because the products fall >20KHz. Yes, this is poor argument for multi-tone signal and IMD.)
Iron transformer this size can be quite flat (and low phase shift) to nearly 20KHz, but (at high impedance) will typically go ringy well before 100KHz. More than enough phase-shift to resonate. We MUST get the NFB well down before this frequency range. The 10KHz-100KHz slope cuts real close to instability.
Another lost detail: the Dynaco *also* has a cap from UL tap to 1st cathode. 390pFd into 47r. What does that do? Exact math is tedious, but say raw gain from 1st cathode to UL tap is 500. (Student should check this.) When reactance of 390pFd is about 47r*500 or 25K, gain falls. Gain is falling by 17KHz! The 82p+18K is almost a red herring; the 390pFd is a powerful compensation and works to (nominally) 9MHz!!
I've seen many such amps be "stable" with all this fluff removed. Sometimes the square-wave response into a dummy load is not bad. However real speakers are not 8.00 Ohms to infinity, and real speech/music may excite ringing which IMs-down into the audio band as a haze. My lawn-engine will "run" without the dashpot on the governor, but under load the speed "hunts" up and down. Lack of outright howling does not mean the system is working cleanly.
As a different argument: Dynaco knew some of what you know and very likely scaled this network for as much fidelity as possible within the limitations of wound iron. I know I would not presume to "better" their work without some different scheme and different trade-offs.
There IS a way to get "flat beyond the audio band". Reduce the gain (and NFB). A triode there would be gain of say 50. The 10KHz corner could be moved (would naturally move) to 40KHz. You have less NFB over the audio band. You have higher THD everywhere speech/music has real power. But hey, it's "flat"!
Whether a particular amp circuit rings, and what frequency it rings at, depends on the details of the build (due to stray capacitance) and the details of the OPT. Hence it may be less than helpful to argue about whether circuit X rings or not; the issue is that circuit X built in layout Y with OPT Z may or may not ring.
What you ask may or may not be possible. Only you can tell, because only you have your amp in front of you.EL34Dave said:
This certainly is a factor. Reducing feedback means that you are likely to need less compensation. You may have greater distortion and higher output impedance, but by reducing feedback you have already decided to accept this.
Like I said at the beginning of this thread, I am a novice hobbyist. I have no training in electronics, but I have no trouble understanding well-explained technical concepts. I do have a signal generator and oscilloscope (both via my computer), so I can generate square waves and view them at any point in the amp circuit. I understand that a dummy resistive load is very different than an actual speaker load, so I also have some cheap test speakers that I don’t mind destroying, if that happens. And I understand that they do not have the same characteristics as my real speakers, but are probably close enough for my purposes (simple tweaking).
I found online a couple suggestions to raise the roll off frequency of the input stage stability network to improve HF in the audio band. I was asking here for a recommendation for the component values in order to cut down on the number of trial and measure steps.
I found online a couple suggestions to raise the roll off frequency of the input stage stability network to improve HF in the audio band. I was asking here for a recommendation for the component values in order to cut down on the number of trial and measure steps.
Dang! I was looking for a way to edit this after posting, oh well. I went back and re-read everyone's responses, which I really do appreciate. My understanding of this technology is improving every day thanks to you guys. Again, I'm not an EE (although I did stay at a Holiday Inn once
). I'm just a tweaker. Am I over my head in all this? Hell yeah! However, I don't intend to dive too much deeper into the details, since I have other hobbies, which also take up time and effort. In fact, this stability network (as well as the cap in the feedback loop) are probably going to be my last tweaks. I know that a valid argument could be made to "just leave well enough alone", but what would be the fun in that?BTW, I'm willing to return the favor if anyone want's to know how to build and tune a Porsche 356, 912, or early 911 road race car. Again, I'm not a ME, but with a lot of research I've been able to be extremely competitive. Kinda like what I'm trying to do with my amp
.
Interesting that you say that. Here's a post on a different forum a number of years ago on that very subject.
Re: Dynaco St-70 feedback loop and cap question - Joe Rosen - Vintage Asylum
I'd suggest you try and work out what you can measure first. I think you say you have a soundcard based oscilloscope and signal generator. So firstly work out what the frequency response is of that setup with a loop-back circuit and report back with a screen grab. Have you made a probe of any kind for the signal in?
The hassle you are likely going to find is that setup will pretty much not be able to generate or display enough bandwidth to be able to 'tweak' and view what is happening where it really matters.
If you want try and tweak, and also start appreciating high frequency response and feedback, then you really should find a cheap analog oscilloscope - preferably two channel with X-Y capability, and some 10:1 probes. You'll also need a cheap signal generator that can do reasonable square waves out to perhaps 20-50kHz, so that say a 5kHz squarewave has neat sharp edges.
The hassle you are likely going to find is that setup will pretty much not be able to generate or display enough bandwidth to be able to 'tweak' and view what is happening where it really matters.
If you want try and tweak, and also start appreciating high frequency response and feedback, then you really should find a cheap analog oscilloscope - preferably two channel with X-Y capability, and some 10:1 probes. You'll also need a cheap signal generator that can do reasonable square waves out to perhaps 20-50kHz, so that say a 5kHz squarewave has neat sharp edges.
The PC scope is a Hantek 6022BE (2 channels, 20 MHz bandwidth). For the signal generator, I'm using the one that comes with TrueRTA, which specs out as 5Hz-48KHz. As TrueRTA is an actual spectrum analyzer, I can use it that capacity once I understand how.
Scope sounds fine, and allows phase change to be assessed (not sure what is in software as manual requires a login, but X-Y will give a reasonable indication as frequency moves to and through unity gain, and you can benchmark original phase change and then check change due to 'tweaking').
I reckon you will need a signal generator other than from a soundcard to do square wave testing, as the soundcard bandwidth may get to 40kHz with a 96kb/s interface, but that is not much bandwidth to assess ringing.
TrueRTA will give good frequency spectrum with a 96kb/s soundcard to assess noise floor and do frequency sweeps out through the audio range, but you will need to revert to scope for any gain bumps beyond about 40kHz (and scope FFT will have poor noise floor).
If you are still keen after getting TrueRTA going, then REW5 is free and allows additional audio amp testing including distortions and THD+N.
What soundcard and probe interface do you have?
I reckon you will need a signal generator other than from a soundcard to do square wave testing, as the soundcard bandwidth may get to 40kHz with a 96kb/s interface, but that is not much bandwidth to assess ringing.
TrueRTA will give good frequency spectrum with a 96kb/s soundcard to assess noise floor and do frequency sweeps out through the audio range, but you will need to revert to scope for any gain bumps beyond about 40kHz (and scope FFT will have poor noise floor).
If you are still keen after getting TrueRTA going, then REW5 is free and allows additional audio amp testing including distortions and THD+N.
What soundcard and probe interface do you have?
The Hantek box is the probe interface. It connects to the PC via USB. I'm using a Behringer UFO202 USB DAC for output, but it's only spec'd out to 20KHz, so I'll probably have to get an actual hardware signal generator in order to test higher freqs. There are some very inexpensive DDS signal generators out there that will go to 65KHz. Would that be sufficient for amp testing?
The square wave input is used to send a wide bandwidth signal through the amp, and any change to the shape of the square wave at the output becomes an indicator of frequency response and stability within the amplifier.
As such, the square wave preferably includes high frequency content that extends beyond the unity-gain bandwidth of the amp and the resonances that could be introduced within the output transformer. That sort of implies a 100+ kHz content, and provides a nice vertical edged square wave as an input signal.
USB based soundcards are great for sinewave testing, and noise floor and harmonic testing as they have huge dynamic range. Your Hantek has relatively poor dynamic range, so the FFT won't be great in comparison to using a soundcard.
However the soundcard needs a line out and line in interface to the amp. That UF0202 has a line out that can drive an amplifier input directly, but the UF0202 line in has a low 27k impedance, and a low max voltage limit, so it needs an interface/probe to connect it to the amplifier output - and that probe needs to provide a flat response out to the soundcard frequency limit. Some USB soundcards provide a 1Meg input option (eg. EMU 0404) or you can modify a $1 soundcard to a 1Meg input, and then use 10:1 scope probe to give good signal voltage capability but there may be some custom compensation needed for flat frequency response.
As such, the square wave preferably includes high frequency content that extends beyond the unity-gain bandwidth of the amp and the resonances that could be introduced within the output transformer. That sort of implies a 100+ kHz content, and provides a nice vertical edged square wave as an input signal.
USB based soundcards are great for sinewave testing, and noise floor and harmonic testing as they have huge dynamic range. Your Hantek has relatively poor dynamic range, so the FFT won't be great in comparison to using a soundcard.
However the soundcard needs a line out and line in interface to the amp. That UF0202 has a line out that can drive an amplifier input directly, but the UF0202 line in has a low 27k impedance, and a low max voltage limit, so it needs an interface/probe to connect it to the amplifier output - and that probe needs to provide a flat response out to the soundcard frequency limit. Some USB soundcards provide a 1Meg input option (eg. EMU 0404) or you can modify a $1 soundcard to a 1Meg input, and then use 10:1 scope probe to give good signal voltage capability but there may be some custom compensation needed for flat frequency response.
I understand that the purity of a square wave in the audio band (particularly with regard to overshoot) can indicate instability at ultrasonic frequencies.
So does that mean that the amp needs to be scoped out to 100+ kHz, or just that the signal generator needs to go to 100+ kHz?
The Hantek accommodates 5V to 20mV, and frequency response from 5000 seconds to 1 nano second. Is that not sufficient?
I have decided to use a hardware signal generator which goes to 65 kHz. Would that do the trick?
The signal generator needs imho bandwidth out beyond 100kHz, which would not be able to be recreated with a USB soundcard sampling rate. That DDS generator, or any digital logic generator, appears fine for square wave generation as it would have a suitably fast rise/fall time no matter what the squarewave frequency was. Normal squarewave test frequency is 1-10kHz.
Your Hantek is fine for viewing the amps response waveform to discern transient overshoot and dampening resonances.
The handtek uses a 10-bit ADC, which limits the dynamic range of an FFT spectrum plot to about 60dB - you will lose some from allowing headroom - that puts distortion harmonics down close to or in the noise floor. A $1 USB adc provides about 100dB range.
Your Hantek is fine for viewing the amps response waveform to discern transient overshoot and dampening resonances.
The handtek uses a 10-bit ADC, which limits the dynamic range of an FFT spectrum plot to about 60dB - you will lose some from allowing headroom - that puts distortion harmonics down close to or in the noise floor. A $1 USB adc provides about 100dB range.
I probably should have mentioned from the beginning, that I am starting with Ned Carlson's (Triode USA Tube Amp Kits Transformers Tubes Dynaco Upgrades and Parts) ST-70 circuit, which already sounds fabulous to me. I'm just trying to tweak, as is my nature. So, I added a 20Hz high pass filter at the input, and slightly lowered the GNFB from 20db to 15db. I'm now just trying to optimize the compensation networks, assuming they need to be modified at all.
I will still test with the sig-gen and the scope, and reply my findings. Again, as a novice at this, any assistance would be welcome.
I will still test with the sig-gen and the scope, and reply my findings. Again, as a novice at this, any assistance would be welcome.
That looks rather like "If I don't understand it then that is because you didn't explain it properly". I'm sure you didn't mean to say this.EL34Dave said:
To understand feedback you first have to understand complex numbers; without that you cannot understand, however perfectly someone explains it to you. Ideally you would also know some complex analysis too, as some aspects of feedback theory are counter-intuitive to those who are baffled by poles and zeros in the complex plane and would not know an analytic function if one dropped on their foot. Fortunately, it is possible to make some progress in using feedback without understanding it.
As I said, it depends on the OPT and the circuit layout details. It also depends on your aim: low HF distortion, good square wave response, HF bandwidth, stability with difficult loads?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.