Dave, thank you for your response. You've left me intrigued 🤔....
excuse my ignorance, but what should I hear if my amp has a high Rout?
Regards
excuse my ignorance, but what should I hear if my amp has a high Rout?
Regards
A peak in frequency response that mimics the shape of the spike in impedance. It will be increased by a fraction of that peak depending on how high the Rout is.
A very high Q peak rising about 10x from base impedance. Might sound like a resonance.
dave
A very high Q peak rising about 10x from base impedance. Might sound like a resonance.
dave
karucho,
Several thoughts on your situation:
1. I think the 6C4C to be a 6.3V filament tube that is essentially a 2A3 (The 6C4C with more filament volts, and less filament amps).
6C4C with 15 Watts plate dissipation, u = 4.1, rp = 760 Ohms, Gm 5400 microMhos.
2. Suppose your output transformer primary is 2500 Ohms, your damping factor will be about 2500/760 = almost 3.3, a little less than that due to the output transformer insertion loss (mostly the primary DCR and secondary DCR losses. Suppose the damping factor is 2.8, that means an 8 Ohm secondary tap would be 8/2.8 = 2.8 Ohms.
3. 2.8 Ohms has to drive the speaker crossover plus drivers that have an impedance of 6 Ohms at 1400Hz, to 68 Ohms at 2800Hz, to 6 Ohms at 10kHz.
4. I am guessing the 68 Ohm peak is from a 6 dB/octave single pole crossover from the mid range to the tweeter.
5. Due to the amplifier's output impedance, its voltage will increase at the 68 Ohms peak. That is what may cause a rise in the acoustic frequency response. But sometimes it is not as bad as it seems, the crossover parts are partially reducing the voltage to the drivers.
6. You could try a series LCR network across the loudspeaker system's input terminals, with the LCR resonance at the same frequency as the 68 Ohm peak (about 2800 Hz).
I am not normally a fan of this, it is resonant (but it is worth trying).
An R of perhaps 20 Ohms could reduce the peak there from 68 Ohms to 20 Ohms.
If Xl and Xc are about 60 Ohms each at 2800Hz, (60 Ohms / 20 Ohms R = 3) that will make the series RLC network effective Q of about 3, which makes the network frequency width about 933Hz, (+/- 467Hz from the 2800Hz center).
3.4mH and 0.95 uF (use 1 uF) are each 60 Ohms at 2800Hz (opposite phase, resonant).
I hope that explains what the series LCR network will do to the original 60 Ohm impedance peak (it had some frequency width of impedance that was larger than 20 Ohms).
A real speaker designer (speaker threads), and many Tubes/Valves readers of this thread, have experience with such networks.
Experts, correct me if I am wrong on my writeup above.
It can not hurt to at least try it, and the series LCR can be outside of the speaker cabinet when you try it.
The price of the parts, the time to try it (listen, not measure), and you find out if you like it or not.
Thanks Everybody!
Have Fun!
Several thoughts on your situation:
1. I think the 6C4C to be a 6.3V filament tube that is essentially a 2A3 (The 6C4C with more filament volts, and less filament amps).
6C4C with 15 Watts plate dissipation, u = 4.1, rp = 760 Ohms, Gm 5400 microMhos.
2. Suppose your output transformer primary is 2500 Ohms, your damping factor will be about 2500/760 = almost 3.3, a little less than that due to the output transformer insertion loss (mostly the primary DCR and secondary DCR losses. Suppose the damping factor is 2.8, that means an 8 Ohm secondary tap would be 8/2.8 = 2.8 Ohms.
3. 2.8 Ohms has to drive the speaker crossover plus drivers that have an impedance of 6 Ohms at 1400Hz, to 68 Ohms at 2800Hz, to 6 Ohms at 10kHz.
4. I am guessing the 68 Ohm peak is from a 6 dB/octave single pole crossover from the mid range to the tweeter.
5. Due to the amplifier's output impedance, its voltage will increase at the 68 Ohms peak. That is what may cause a rise in the acoustic frequency response. But sometimes it is not as bad as it seems, the crossover parts are partially reducing the voltage to the drivers.
6. You could try a series LCR network across the loudspeaker system's input terminals, with the LCR resonance at the same frequency as the 68 Ohm peak (about 2800 Hz).
I am not normally a fan of this, it is resonant (but it is worth trying).
An R of perhaps 20 Ohms could reduce the peak there from 68 Ohms to 20 Ohms.
If Xl and Xc are about 60 Ohms each at 2800Hz, (60 Ohms / 20 Ohms R = 3) that will make the series RLC network effective Q of about 3, which makes the network frequency width about 933Hz, (+/- 467Hz from the 2800Hz center).
3.4mH and 0.95 uF (use 1 uF) are each 60 Ohms at 2800Hz (opposite phase, resonant).
I hope that explains what the series LCR network will do to the original 60 Ohm impedance peak (it had some frequency width of impedance that was larger than 20 Ohms).
A real speaker designer (speaker threads), and many Tubes/Valves readers of this thread, have experience with such networks.
Experts, correct me if I am wrong on my writeup above.
It can not hurt to at least try it, and the series LCR can be outside of the speaker cabinet when you try it.
The price of the parts, the time to try it (listen, not measure), and you find out if you like it or not.
Thanks Everybody!
Have Fun!
Last edited:
Dave, thanks again. I haven't tried the system yet, but when I do I'll pay attention to "Might sound like a resonance" đź‘Ť.
6A3sUMMER, an enlightening technical explanation. The output transformer primary is 2500 Ohms, it is the Hammond 1627SEA.
When I start up the set, I will have a place to start if I detect any resonance.
Thank you very much ;-)
Regards
6A3sUMMER, an enlightening technical explanation. The output transformer primary is 2500 Ohms, it is the Hammond 1627SEA.
When I start up the set, I will have a place to start if I detect any resonance.
Thank you very much ;-)
Regards
Hi folks,
Sorry if I've stirred things up, didn't mean to; my purpose was to show, for the purposes of discussing speakers elsewhere, that my amp was engineered to give more real world drive ability than one would expect from 3.5 watts; this is the case, though I know you experts have gone far further.
To give the cct etc means diving into my archives, not easy at present.
3.5 watts (not measured) from parallel 45's; output impedance on 4 ohm setting 1.9 ohms, measured. Conventional output stage, direct heated, good OPT. PSUs described above, low C and low DCR chokes.
Now to address some issues above.
My experience with added ccts for Z correction seems to agree with Chris Holbeck's comment that it may not be good; somehow, it seemed to take some of the life out of the music (and this system is fast, direct, musical). Of course I am short on watts, perhaps some extra loss mattered. Others will find different!
The speakers were designed allowing for the amp Zout, in both bass alignment and crossover design; then they were measured, finding a slight dip and peak near the crossover freq and the crossover was modified to correct these; at the same time, BSC and tweeter level were adjusted, these mainly by ear. I tend to think that anomalies due to varying Z were sorted out in these processes, without me knowing much about it.
A general thought which may be wrong! (I like brainstorming; never suppress a thought because it may be wrong).
An impedance peak in a speaker may be due to mechanical resonance; if so, less power will be needed at that frequency for a given cone movement?
Due to the high speaker impedance, less power will be delivered due to high Zout of a SET?
So, is the situation to some extent self cancelling?
Karucho, my experience with 3.5 watts (giving good room-filling volume without stress in my 18ft x 14ft room) suggests you will be very much OK. Bass performance is irrelevant, but you may need to adjust the crossover because of amp Zout. I can't help here, other than to say this. Modelling of crossovers is very good and can allow for amp Zout, but it's unlikely to be 100% right and you may well need to measure the speakers, and use ears, to make final adjustments. Well worth doing. Experts in the multi way speaker forum should be able to help with all this.
Good discussions guys even though most of it is above my head. 🙂
Sorry if I've stirred things up, didn't mean to; my purpose was to show, for the purposes of discussing speakers elsewhere, that my amp was engineered to give more real world drive ability than one would expect from 3.5 watts; this is the case, though I know you experts have gone far further.
To give the cct etc means diving into my archives, not easy at present.
3.5 watts (not measured) from parallel 45's; output impedance on 4 ohm setting 1.9 ohms, measured. Conventional output stage, direct heated, good OPT. PSUs described above, low C and low DCR chokes.
Now to address some issues above.
My experience with added ccts for Z correction seems to agree with Chris Holbeck's comment that it may not be good; somehow, it seemed to take some of the life out of the music (and this system is fast, direct, musical). Of course I am short on watts, perhaps some extra loss mattered. Others will find different!
The speakers were designed allowing for the amp Zout, in both bass alignment and crossover design; then they were measured, finding a slight dip and peak near the crossover freq and the crossover was modified to correct these; at the same time, BSC and tweeter level were adjusted, these mainly by ear. I tend to think that anomalies due to varying Z were sorted out in these processes, without me knowing much about it.
A general thought which may be wrong! (I like brainstorming; never suppress a thought because it may be wrong).
An impedance peak in a speaker may be due to mechanical resonance; if so, less power will be needed at that frequency for a given cone movement?
Due to the high speaker impedance, less power will be delivered due to high Zout of a SET?
So, is the situation to some extent self cancelling?
Karucho, my experience with 3.5 watts (giving good room-filling volume without stress in my 18ft x 14ft room) suggests you will be very much OK. Bass performance is irrelevant, but you may need to adjust the crossover because of amp Zout. I can't help here, other than to say this. Modelling of crossovers is very good and can allow for amp Zout, but it's unlikely to be 100% right and you may well need to measure the speakers, and use ears, to make final adjustments. Well worth doing. Experts in the multi way speaker forum should be able to help with all this.
Good discussions guys even though most of it is above my head. 🙂
Mike C,
Thanks for your impressions. Yes, it seems that 4.5w will be enough, in my room, similar in size to yours.
I have decided that before worrying, I will wait and listen to the system very carefully, to see what I detect...fingers crossed 🙂
Regards
Thanks for your impressions. Yes, it seems that 4.5w will be enough, in my room, similar in size to yours.
I have decided that before worrying, I will wait and listen to the system very carefully, to see what I detect...fingers crossed 🙂
Regards
Is that the equivalent od being able to deliver significantly more current than one would expect from the Power = VA. (forgetting the cos(pahse angle) term?
ie what you look for in an SS ampliifetr that will drive very capacitive ESLs.
dave
Last edited:
The problem with impedance peak (either from driver or crossover) is not just reduced power at resonance frequency. At such peak, the speaker presents capacitive (on one slope) or inductive (on the other slope) load to amplifier output. Such reactive loads cause elliptic load lines associated with additional distortion.
Mike C,
You said:
"My experience with added ccts for Z correction seems to agree with Chris Holbeck's comment that it may not be good; somehow, it seemed to take some of the life out of the music (and this system is fast, direct, musical)"
I agree with you and Chris.
That often is true.
It is a compromise, and the listening results of extra networks is very dependent on both the speaker & crossover, and on the amplifier.
In my Post # 43, I picked a 20 Ohm resistor for the LCR network, and capacitor and inductor values to give a Q of about 3.
That gives a moderate amount of change of the impedance at and around 2800Hz.
The low Q across the loudspeaker terminals, is less likely to cause a ringing problem.
It also only reduces the efficiency of the loudspeaker.
I tried to use the network somewhat conservatively, and not to squash the impedance bump completely.
Lower resistance, 10 Ohms for example, would flatten the impedance nicely, but would also . . .
Reduce the loudspeaker efficiency in a fair size bandwidth centered around the 2800 Hz frequencies,
And it might be more likely to cause ringing
Too much of a good thing is not always a good thing (it is probably what sucks the life out so many systems).
All things in Moderation
As always, Have Fun!
It is a Mean thing to say that Karl Friedrich Gauss was an Average man, who was Centered at the Peak of his Gaussian Curve.
You said:
"My experience with added ccts for Z correction seems to agree with Chris Holbeck's comment that it may not be good; somehow, it seemed to take some of the life out of the music (and this system is fast, direct, musical)"
I agree with you and Chris.
That often is true.
It is a compromise, and the listening results of extra networks is very dependent on both the speaker & crossover, and on the amplifier.
In my Post # 43, I picked a 20 Ohm resistor for the LCR network, and capacitor and inductor values to give a Q of about 3.
That gives a moderate amount of change of the impedance at and around 2800Hz.
The low Q across the loudspeaker terminals, is less likely to cause a ringing problem.
It also only reduces the efficiency of the loudspeaker.
I tried to use the network somewhat conservatively, and not to squash the impedance bump completely.
Lower resistance, 10 Ohms for example, would flatten the impedance nicely, but would also . . .
Reduce the loudspeaker efficiency in a fair size bandwidth centered around the 2800 Hz frequencies,
And it might be more likely to cause ringing
Too much of a good thing is not always a good thing (it is probably what sucks the life out so many systems).
All things in Moderation
As always, Have Fun!
It is a Mean thing to say that Karl Friedrich Gauss was an Average man, who was Centered at the Peak of his Gaussian Curve.
I felt that way until finally sorting out my driver. The current topology is two stage using a Pimm-style CCS loaded pentode with an optimized mosfet follower as a driver capable of 120 Vp-p at 0.02% THD, predominantly H2 and H3. The 828 UL output stage requires 60x gain to reach grid positive operation. The driver stage gain is roughly 600, leaving plenty left for global feedback.
The end result is a ~25 watt SE amp with a good DF and a low and very benign distortion profile capable of 20 watts at 30 Hz well below 1% THD. Handles any musical genre well but GNFB demanded the most linear possible driver in my experience for the amp to sound its best. It's not a 60's style loop-around-all-distortions-and-forget panacea.
rdf: Sounds like an interesting amp, is there a schematic or even a thread about it somewhere?
My comment about global NFB was aimed at the typical "connect a resistor from the speaker tap to the input tube's cathode and adjust until it sounds like a castrated transistor radio but measures good" kind of feedback.
Due to my lack of impulse control I'm currently collecting parts for a 300B SET even though I know from experience that my speakers prefer amps with a bit more damping than what the average SET provides. Or, to rephrase it: I have a pair of big expensive OPTs (LL1688) left from an ambitious transmitter amp project that never happened and decided to wire them for a lower impedance ratio and use 300B instead of 813/GM70 running at absolutely lethal voltages.
To match my speakers better, I might have to come up with a way to increase the damping factor a bit compared to what I'd get from a zero NFB 300B SET.
My comment about global NFB was aimed at the typical "connect a resistor from the speaker tap to the input tube's cathode and adjust until it sounds like a castrated transistor radio but measures good" kind of feedback.
Due to my lack of impulse control I'm currently collecting parts for a 300B SET even though I know from experience that my speakers prefer amps with a bit more damping than what the average SET provides. Or, to rephrase it: I have a pair of big expensive OPTs (LL1688) left from an ambitious transmitter amp project that never happened and decided to wire them for a lower impedance ratio and use 300B instead of 813/GM70 running at absolutely lethal voltages.
To match my speakers better, I might have to come up with a way to increase the damping factor a bit compared to what I'd get from a zero NFB 300B SET.
The Altec 816-lookalikes and the white horns on top of them. Sensitivity is surely high enough but the crossovers are a bit complex and the "hornflex" cabs seem a bit picky about the source impedance:
Nice looking horns, how low do they go? How high?
What does the impedance look like?
I could see issues with the big reflexes. The alternative of decreasing the Rout would be to flatten its impedance peak by pushing it aperiodic. Or biamp. Or triamp.
dave
What does the impedance look like?
I could see issues with the big reflexes. The alternative of decreasing the Rout would be to flatten its impedance peak by pushing it aperiodic. Or biamp. Or triamp.
dave
Thanks!
The white mid horns are 200Hc LeCleach horns, hi passed at 400Hz 24dB/octave. With a bit of passive EQ they reach 12k or so, where the supertweeters take over.
The hornflexes are a bit of a mess, designed to get maximum LF output from the ported back chambers and miminum horn gain to achieve a reasonably flat response down to 50Hz or so. Works fairly well but the midbass is a bit pronounced even with BSC filters involved. The impedance is probably very far from flat, so biamping with SS amps below 400Hz would probably be a good idea.
The white mid horns are 200Hc LeCleach horns, hi passed at 400Hz 24dB/octave. With a bit of passive EQ they reach 12k or so, where the supertweeters take over.
The hornflexes are a bit of a mess, designed to get maximum LF output from the ported back chambers and miminum horn gain to achieve a reasonably flat response down to 50Hz or so. Works fairly well but the midbass is a bit pronounced even with BSC filters involved. The impedance is probably very far from flat, so biamping with SS amps below 400Hz would probably be a good idea.
Not yet. The design is still in flux. After last posting a revisit of the output bias points and loading resulted in ~35 watts with under ~0.6% distortion. When the analysis is complete there is intent to start a thread on the modified follower. The comment on GNFB is understood but my sense is the receiver sound is related to treating it as a universal fix. Driver linearity and drive ability appear to be much more important with GNFB than in open loop circuits. It certainly sounds that way to me.
We appear to be on similar paths. The intent this afternoon is swapping out the 828 for the more G2 robust 813 and go for 50 watts of universal, low distortion SE. Also back of mind for the next project is a no-FB, transformer coupled 300B SE.
Edit: BTW 35 watts was with a moderate 850 Vdc B+ and under 60 watts plate dissipation, depending on what you consider lethal.
rdf: I'm looking forward to that thread!
This thread by SpreadSpectrum shows an interesting way to apply large amounts of negative feedback in SE amps:
https://www.diyaudio.com/community/...ow-distortion-a2-dht-se-amp-prototype.357559/
As I mentioned earlier, I've had some luck with adding CFB (cathode feedback) in my output stages. Doing this the usual way in a 300B SET with 5k/8R OPTs would result in something like 1dB of local feedback, hardly worth the trouble. Unlike most OPTs, my large Lundahls have two primary windings that are supposed to be externally connected in series to make the full winding. This could allow the design of a split load output stage with 50% CFB.
Assuming it even works as intended, it should result in an output stage with low distortion and low Zout, but the demands on the driver stage would be ridiculous (200Vrms or more for full power).
This thread by SpreadSpectrum shows an interesting way to apply large amounts of negative feedback in SE amps:
https://www.diyaudio.com/community/...ow-distortion-a2-dht-se-amp-prototype.357559/
As I mentioned earlier, I've had some luck with adding CFB (cathode feedback) in my output stages. Doing this the usual way in a 300B SET with 5k/8R OPTs would result in something like 1dB of local feedback, hardly worth the trouble. Unlike most OPTs, my large Lundahls have two primary windings that are supposed to be externally connected in series to make the full winding. This could allow the design of a split load output stage with 50% CFB.
Assuming it even works as intended, it should result in an output stage with low distortion and low Zout, but the demands on the driver stage would be ridiculous (200Vrms or more for full power).