By this I mean for less than ideal speakers. Due to WAF factor, my MTMs of about 91dB are less than ideal but results are good.
I'm currently discussing making speakers for SETs, seems equally valid to discuss making SETs for normal speakers here, they are intimately linked.
I will cover the principles here, if anyone wants the amp design I can look it up and post it; the principles should apply to any SET.
A critical friend (I liken him to Craig Revel Horwood, he is fair but harsh for my own good) is highly into dynamics, with a powerful active Naim setup.
He has said of my present amp/speaker system that it has fast clean leading edges, and that at last I have some dynamics. Funnily , this is the least powerful amp I have even had; I've had EL34 PP and 2 300B SETs.
This amp is 45 parallel SET, 6C8G/E83F mu stage driver. Yes, I know parallel 45 is much like a 2A3, but I heard a 45 and wanted 45's, did it, right or wrong. I was thinking of 45 PP but a friend, expert, who knows me well said Mike, you're a SET guy, go parallel.
For dynamics, my principles were:
It's all in the PSU (well, very very nearly all).
I tend to think this; compare say a 3watt amp with 300 watt PSU, vs a 30 watt amp with 30 watt PSU; we know which will measure more powerful but I'm sure I know which will sound more powerful in the real world driving speakers.
Big mains TX, over rated by about 3 times.
Use a separate PSU for the driver stage. I learnt this from Naim, who split PSUs by stage rather than by channel; it prevents the sag from a drop in output stage PSU from getting to the driver stage; definitely recommended.
I use a GZ34 for the driver stage, but for the output stage a pair of 6CJ3 damper diodes in a full wave (CT transformer); very low voltage drop.
Both PSUs are 2 pi; low DCR chokes in the output stage PSU. Film caps including motor run oil caps, medium values; I forget exactly but of the order of 5uF, 20uF, 40uF.
In particular, the output stage was tuned in PSUD for fast clean transient response, using the PSUD step load function. Search in Audio Asylum. Tube DIY Asylum to learn how to do this; it really works, hence the fast clean leading edges, good 'speed' to the sound generally; nothing slow or lazy to this amp.
This is my approach, other ways are equally valid and may well be better.
For good dynamics to drive real speakers, a good driver stage is vital. To get real power and dynamics I like to be able to swing almost twice the bias voltage each way from a low impedance source. Many ways to do this; I chose a mu stage as I've had good results.
6SL7 does not have enough current; 6SN7 does not quite have enough gain; I needed a 6SM7 but there never was one.
6C8G is the 6SM7 that never was! Mid way between L7 and N7. Runs well at 5mA, and the E83F pentode runs at 12mA total current and is a sweet sounding pentode, I use it a lot as active load.
To run these valves at good operating points, giving say 90V swing both ways, needs a lot of B+, well over 400V if I remember; another reason for the separate PSU for the driver.
Anyway, with a low Zout, this driver will be able to drive the triodes well, giving a lot of current when it's called for.
Which leads to the next issue. Driving the triodes hard (over driving them?) will lead to grid current, which can be a problem. The fix is to use a grid choke, highly recommended.
The result of all this is a decent ability to drive speakers; even at half the power of my 300B amp, it drives them much better.
I'm not saying these things are the only way! You experts will have found good ways to do similar. But I can recommend these principles; in particular, PSUD step response modelling.
I'm currently discussing making speakers for SETs, seems equally valid to discuss making SETs for normal speakers here, they are intimately linked.
I will cover the principles here, if anyone wants the amp design I can look it up and post it; the principles should apply to any SET.
A critical friend (I liken him to Craig Revel Horwood, he is fair but harsh for my own good) is highly into dynamics, with a powerful active Naim setup.
He has said of my present amp/speaker system that it has fast clean leading edges, and that at last I have some dynamics. Funnily , this is the least powerful amp I have even had; I've had EL34 PP and 2 300B SETs.
This amp is 45 parallel SET, 6C8G/E83F mu stage driver. Yes, I know parallel 45 is much like a 2A3, but I heard a 45 and wanted 45's, did it, right or wrong. I was thinking of 45 PP but a friend, expert, who knows me well said Mike, you're a SET guy, go parallel.
For dynamics, my principles were:
It's all in the PSU (well, very very nearly all).
I tend to think this; compare say a 3watt amp with 300 watt PSU, vs a 30 watt amp with 30 watt PSU; we know which will measure more powerful but I'm sure I know which will sound more powerful in the real world driving speakers.
Big mains TX, over rated by about 3 times.
Use a separate PSU for the driver stage. I learnt this from Naim, who split PSUs by stage rather than by channel; it prevents the sag from a drop in output stage PSU from getting to the driver stage; definitely recommended.
I use a GZ34 for the driver stage, but for the output stage a pair of 6CJ3 damper diodes in a full wave (CT transformer); very low voltage drop.
Both PSUs are 2 pi; low DCR chokes in the output stage PSU. Film caps including motor run oil caps, medium values; I forget exactly but of the order of 5uF, 20uF, 40uF.
In particular, the output stage was tuned in PSUD for fast clean transient response, using the PSUD step load function. Search in Audio Asylum. Tube DIY Asylum to learn how to do this; it really works, hence the fast clean leading edges, good 'speed' to the sound generally; nothing slow or lazy to this amp.
This is my approach, other ways are equally valid and may well be better.
For good dynamics to drive real speakers, a good driver stage is vital. To get real power and dynamics I like to be able to swing almost twice the bias voltage each way from a low impedance source. Many ways to do this; I chose a mu stage as I've had good results.
6SL7 does not have enough current; 6SN7 does not quite have enough gain; I needed a 6SM7 but there never was one.
6C8G is the 6SM7 that never was! Mid way between L7 and N7. Runs well at 5mA, and the E83F pentode runs at 12mA total current and is a sweet sounding pentode, I use it a lot as active load.
To run these valves at good operating points, giving say 90V swing both ways, needs a lot of B+, well over 400V if I remember; another reason for the separate PSU for the driver.
Anyway, with a low Zout, this driver will be able to drive the triodes well, giving a lot of current when it's called for.
Which leads to the next issue. Driving the triodes hard (over driving them?) will lead to grid current, which can be a problem. The fix is to use a grid choke, highly recommended.
The result of all this is a decent ability to drive speakers; even at half the power of my 300B amp, it drives them much better.
I'm not saying these things are the only way! You experts will have found good ways to do similar. But I can recommend these principles; in particular, PSUD step response modelling.
I've been using SE amps for many years now. Like you I use a substantial external PSU and a separate PSU for the driver. I have pretty average speakers - Mission 761 - nothing high efficiency, but smallish room and I never listen loud. So I get away with 2 stages in the amp. For my purposes I need about 18W dissipation. 15W works (e.g. a 2a3) but 18W is better. I started in the conventional way with 2a3 and 300b but was never satisfied with the driver tube, being used to DHTs. And I found the driver tube had a big effect on the final sound. My breakthrough was to "invert" the DHT and put it first. So I used a 26 or 10Y. It then followed that I needed output tubes with around 18x gain, i.e mu=18. So I've been using EL12n in triode which are cheap in Europe (use 7591 in the USA).
I'm interested that you found your best option in PSE with 2 tubes in parallel. My latest high gain output stage is parallel E80L in triode, used in preference to EL84 because they are SQ, gold pins, and sound great. You read on several threads that you should prefer one single high power tube to two lower power tubes in parallel. I don't know how much truth is in that, but I'm getting a very nice sound with the E80Ls. Like you with your 45s, you can find some marvellous tubes that need to be used in parallel as outputs.
Interested in others' opinions here.
I'm interested that you found your best option in PSE with 2 tubes in parallel. My latest high gain output stage is parallel E80L in triode, used in preference to EL84 because they are SQ, gold pins, and sound great. You read on several threads that you should prefer one single high power tube to two lower power tubes in parallel. I don't know how much truth is in that, but I'm getting a very nice sound with the E80Ls. Like you with your 45s, you can find some marvellous tubes that need to be used in parallel as outputs.
Interested in others' opinions here.
Last edited:
Well, in my experience it's nice to have brute power output in an SET because that allows one to trade some voltage swing for driving current using a stepdown autoformer such as the Spelz Zero. Of course, overbuilt amp power supplies are a must.
Personally, I drive Infinity RSIIb speakers with my 200WPC SET in this manner. The addition of the Speltz in 4:1 step-down really gave the amps a hammer lock on the speaker drivers. A solution worth trying if you have power to spare.
Personally, I drive Infinity RSIIb speakers with my 200WPC SET in this manner. The addition of the Speltz in 4:1 step-down really gave the amps a hammer lock on the speaker drivers. A solution worth trying if you have power to spare.
Attachments
Lots of great ideas here. I want to try the 26 sometime - I use the 27 in my best preamp (mu stage with guess what, E83F pentode), and I've been told about the 10Y.
Many (most) normal loudspeakers — let’s say a cone+dome — have impedance curves that do not mate well with amplifiers — like most SETs — have a highish Rout and the impedance curve imposes itself on the system FR. More important than sensitivity to my mind,
A typical SET needs a speaker designed for using with its “peculiarities”.
this is why you see many callling SETs as amps with built in (fixed) tone control ignoring what a SET can do so well.
Doesn’t help much if the speaker is not matched to the SET. I suspect to get the Rout low enuff to work with a “norma” loudspeaker one would have to use large amounts of feedback.
dave
You just think so!
Try graphite anode 801 .... IT's addictive!
But nowadays you need to sell your kidney for a good pair. :-(
Am I right in thinking that the potential weakness in matching a SE amp to a speaker is the bass response? So can speaker experts like Dave give us an idea of where the matching becomes important - 100Hz, 300Hz, 500Hz.........?
Given that bi-amping and using subs is becoming more common, at what frequency does a SE amp start to show its best qualities?
Given that bi-amping and using subs is becoming more common, at what frequency does a SE amp start to show its best qualities?
Bass resonance is an issue. But impedance issues higher up cause issues that are probably more of an issue. As Mike has shown, the output impedance can largely be dealt with in a reflex.
If the impedance response is not relatively flat it will always impose itself on the music.
For example a pretty typical impedance for a Klipsch (often recommended for SETs due to specified sensitivity (seems in practice to be exagerated about 3dB).
This loudspeaker mated with a SET with high Rout will have elevated response above about 1kHz due to the large rise in impedance. And since most of thes ewill be used with SS amps with low Rout so the bass will probably sound “loose & wooly"
dave
Last edited:
<< As Mike has shown, the output impedance can largely be dealt with in a reflex.
You mean a ported speaker? Could you go into more detail - sounds important.
You mean a ported speaker? Could you go into more detail - sounds important.
Andy: https://www.diyaudio.com/archive/ar...hly-modern-tube-phono-preamp-prev-thread.html
(Morgan Jones article about set and speakers for them)
(Morgan Jones article about set and speakers for them)
"Many (most) normal loudspeakers — let’s say a cone+dome — have impedance curves that do not mate well with amplifiers — like most SETs — have a highish Rout and the impedance curve imposes itself on the system FR. More important than sensitivity to my mind,"
Nice points Dave. Expanding on your point, as the power output increases, the minimum signal plate current of the output tube decreases
and clips at zero plate current. Also, as the signal plate current drops, the plate resistance of the output tube increases, and
is non linear in most, if not all cases.
As the plate resistance increases, the output impedance (Z) of the amplifier increases. This reduces the damping factor. Hence the damping factor
varies as the plate current varies with the musical signal. As the power output is increased, the damping factor varies more and more, approaching no damping/control of the cone of each driver.
Attached is a set of curves of a typical triode. Disregard the X, Y numbers. Instead notice the rp curvature of the plate resistance as the plate
current varies. That causes distortion.
cheers and hope this helps in understanding.
pos
Nice points Dave. Expanding on your point, as the power output increases, the minimum signal plate current of the output tube decreases
and clips at zero plate current. Also, as the signal plate current drops, the plate resistance of the output tube increases, and
is non linear in most, if not all cases.
As the plate resistance increases, the output impedance (Z) of the amplifier increases. This reduces the damping factor. Hence the damping factor
varies as the plate current varies with the musical signal. As the power output is increased, the damping factor varies more and more, approaching no damping/control of the cone of each driver.
Attached is a set of curves of a typical triode. Disregard the X, Y numbers. Instead notice the rp curvature of the plate resistance as the plate
current varies. That causes distortion.
cheers and hope this helps in understanding.
pos
This set of relexes from Scott is illustrative.
http://wodendesign.com/downloads/King-of-Swingers.pdf
dave
http://wodendesign.com/downloads/King-of-Swingers.pdf
dave
If you’re willing to run tetrodes in triode mode look at the KT120 and 150. I’m loving the sound of them and they are easy to drive. Just an idea!
First, nice to see you here Andy! We've corresponded in the past.
Next, thanks for the interest from all you folks, you are way ahead of me in terms of expertise, I'm just a bit of a diy hacker.
Some things raised here are addressed (to some extent) in multi way speakers:
SET Friendly MTM, Seas ER18RNX and 27TDFC
Briefly, the high amp Zout raises the Qts very significantly, making it hard or impossible to arrange a normal reflex bass alignment. But, within my constraint of only moderate box volume, WINisd modelling allowed a bass hump to be minimised to a low level while keeping group delay within a limit specified by experts. Done by a lower tuning of the port, significantly lower than Fs. As I listen at moderate levels, port misbehaviour has not been a problem. Results then are perhaps not ideal, but highly acceptable to me. I'm lucky.
As for impedance variations; I tried various zobels and didn't like any of them. I forget now but I think it may have been one on a drive unit, another across the whole crossover. Just seemed to mess up the sound quality; and measured FR was not bad at all, and the sound seemed balanced. I've been lucky.
Next, thanks for the interest from all you folks, you are way ahead of me in terms of expertise, I'm just a bit of a diy hacker.
Some things raised here are addressed (to some extent) in multi way speakers:
SET Friendly MTM, Seas ER18RNX and 27TDFC
Briefly, the high amp Zout raises the Qts very significantly, making it hard or impossible to arrange a normal reflex bass alignment. But, within my constraint of only moderate box volume, WINisd modelling allowed a bass hump to be minimised to a low level while keeping group delay within a limit specified by experts. Done by a lower tuning of the port, significantly lower than Fs. As I listen at moderate levels, port misbehaviour has not been a problem. Results then are perhaps not ideal, but highly acceptable to me. I'm lucky.
As for impedance variations; I tried various zobels and didn't like any of them. I forget now but I think it may have been one on a drive unit, another across the whole crossover. Just seemed to mess up the sound quality; and measured FR was not bad at all, and the sound seemed balanced. I've been lucky.
Last edited by a moderator:
My recipe for more complex speaker loads is to use low Rp tubes (triode wired sweep tubes) and a good amount of CFB. Subjectively, such amps can handle difficult loads quite well without resorting to global NFB.
Thanks Fuling!
Do you have an example? Just thinking about required drive levels for these, triode wired sweep tubes have low gain, and CFB on top of that.
Do you have an example? Just thinking about required drive levels for these, triode wired sweep tubes have low gain, and CFB on top of that.