Will GM (transconductance) changes affect G1 current requirements? Higher GM means higher gain, therefore less G1 I needed for max output?
Last edited:
That was my next Q-
Could 7591 instead of 6L6 in prev shown design use low mu triode as phase splitter instead of diff driver setup like ST70?
Tom Bavis;6824341The "super" ST-70's output transformer would saturate at the same voltage as in a normal ST-70 - which would be twice the power. Yes said:
Acrylated silicone resins have excellent barrier (O2) protection and are quite tolerant of high bulb temperatures (unlike other UV curing resins which are sensitive to heat). They can be applied such they completely envelop the bulb. Curing via UV or EB would be entirely conformal.
Topic for another thread I believe. It's a pity that 7868's go dodo so quickly.
I think you're referring to class A2, AB2, or B2 power amplifiers? Well, you need to compare datasheets to answer this. Anyway, as transconductane is the grid voltage to plate current variation relationship, I don't see any reason why it should.
Depends on the phase splitter. LPT's and paraphases benefit from high µ tubes (at least if there's no CCS tail...)
Best regards!
sorry the transconductance of the *HB5 doesn't come even close to the 8417, although clearly the anode structure looks pretty similar. (9100mhos).
Even semi dead 8417 hit around 18,000, and proper new condition ones are up right up there around 22000-24000.
I used the rare and expensive Toshiba xG-Bx as a substitute which seems to manage KT88 levels of HT and Pa (I get them to do 43W with no sign of red).
No worries at 700V and I reckon they can take more, despite it being totally undocumented in the Jap language docs.
Sadly if you use the Tosh tube in one channel of my "heavy metal" monoblocs compared with my other channel using Sylv 8417, there's a quite a significant difference in gain.
There are 2 types of glass envelope, one of them being pretty tall so that can cause space issues.
Then, when you use the OPT as cathode feedback (CT secondaries is great!), you need all the gain you can get, cos it doubles drive requirements.
In those amps, even a quad of 7591 would not cut it, (the amp was made for PPP using 8 x 8417!).
Having said that, there's something about those old "westy" US designed valves, that adds some magic to the sound, and I am not alone in saying so.
Curves from the manufacturer are not everything, something in the linearity department happens to make them have that peculiarly likeable sound signature....
(and I can listen to mine for hours and hours with zero listener fatigue!)
Attachments
Last edited:
re: Kodabmx
I think the grid 1 gm would be the same for pentode and triode -IF- the "plate" V is held constant.
The reduced gm often shown for triodes apparently is for specific operating conditions (specified load R) where the plate V changes are counteracting the grid1 V change to some extent. A CCS load would theoretically reduce the grid 1 gm all the way to zero, by totally counteracting the grid1 effect, leaving constant current.
So for a "trioded" pentode, one would use the internal Mu to calc. an effective gm2 from gm1. gm2 = gm1 x 1/mu. Then use the voltage variation with the specified "plate" load to calc. the effective counter gm2 current effect to subtract from gm1 current effect. (all of which depends on plate current level as well! ie, gm variation with approx. SQRT of current)
I think the grid 1 gm would be the same for pentode and triode -IF- the "plate" V is held constant.
The reduced gm often shown for triodes apparently is for specific operating conditions (specified load R) where the plate V changes are counteracting the grid1 V change to some extent. A CCS load would theoretically reduce the grid 1 gm all the way to zero, by totally counteracting the grid1 effect, leaving constant current.
So for a "trioded" pentode, one would use the internal Mu to calc. an effective gm2 from gm1. gm2 = gm1 x 1/mu. Then use the voltage variation with the specified "plate" load to calc. the effective counter gm2 current effect to subtract from gm1 current effect. (all of which depends on plate current level as well! ie, gm variation with approx. SQRT of current)
Last edited:
9100 is the 21HB5 gm at 46 mA. I was previously calculating the increased gm of the 21HB5 (10200) at the 60 mA level quoted (on datasheet) for the 7591.
So the 21HB5 does as well as the 7591/7868 (in parallel) of mimicking the 8417. But they do all fall a little short of the of the 23000 gm of the 8417 at 100 mA.
8417 and especially the Toshiba tube are pretty scarce these days.
25HX5 (another Japanese tube) will get you 14000 gm per tube, but falls short on most other parameters for subbing as a 7591 or in parallel to make an 8417.
You could use 3x 12HL7 in parallel to get 60000 gm at 30 Watts. Stability? It was another $1 list tube, like the 21HB5.
The 21HB5 has an identical 35 mm long plate as the 8417. Its 18 Watt rating for TV Horiz. deflection is quite conservative. Most 24 Watt TV sweep tubes have the 35 mm plate. IBM used to use 21HB5 tubes for their early VT computer monitors. -RELIABLE-
With the low 50V knee Voltage, the 21HB5 is more efficient. It has better triode curves (better sound due to less dist. from forced gm) With two in parallel to get the high gm model, it has more than 2x the current capability of the 8417. And with a truck-load of $1 tubes available.... well you get the idea. Two sockets are better than one. (With a curve tracer for matching.) 50 or 60 Watts from a single pair is usually sufficient anyway.
So the 21HB5 does as well as the 7591/7868 (in parallel) of mimicking the 8417. But they do all fall a little short of the of the 23000 gm of the 8417 at 100 mA.
8417 and especially the Toshiba tube are pretty scarce these days.
25HX5 (another Japanese tube) will get you 14000 gm per tube, but falls short on most other parameters for subbing as a 7591 or in parallel to make an 8417.
You could use 3x 12HL7 in parallel to get 60000 gm at 30 Watts. Stability? It was another $1 list tube, like the 21HB5.
The 21HB5 has an identical 35 mm long plate as the 8417. Its 18 Watt rating for TV Horiz. deflection is quite conservative. Most 24 Watt TV sweep tubes have the 35 mm plate. IBM used to use 21HB5 tubes for their early VT computer monitors. -RELIABLE-
With the low 50V knee Voltage, the 21HB5 is more efficient. It has better triode curves (better sound due to less dist. from forced gm) With two in parallel to get the high gm model, it has more than 2x the current capability of the 8417. And with a truck-load of $1 tubes available.... well you get the idea. Two sockets are better than one. (With a curve tracer for matching.) 50 or 60 Watts from a single pair is usually sufficient anyway.
Last edited:
Pentode Transconductance is defined as what happens when:
Plate Voltage held constant
Screen Voltage held constant
G1 Bias Voltage is varied (Delta Bias)
Measure the change in Plate Current (Delta Plate Current)
Transconductance, Gm = Delta Plate Current / Delta Bias Voltage
Example:
Initial setting
Bias voltage is -25V
Plate current is 40mA
Change Bias voltage to -24V
Plate current is now 50mA
Delta Plate current is 10mA (0.010A)
Delta Bias is 1V
Transconductance: 0.010A / 1V = 10,000 uMhos.
Now repeat, but change the bias voltage in the other direction:
Initial setting
Bias voltage is -25V
Plate current is 40mA
Change Bias voltage to -26V
Plate current is now 31mA
Delta Plate current is 9mA (0.009A)
Delta Bias is 1V
Transconductance: 0.009A / 1V = 9,000 uMhos.
Notice that the transconductance is different for the 1V delta bias voltage, versus the 1V delta bias voltage change in the opposite direction.
That difference in transconductance is what causes 2nd harmonic distortion, and causes 2nd order intermodulation distortion.
Also, Beam Power Tube Transconductance is defined and measured in the same way.
Does anybody need a simple schematic of that test?
Plate Voltage held constant
Screen Voltage held constant
G1 Bias Voltage is varied (Delta Bias)
Measure the change in Plate Current (Delta Plate Current)
Transconductance, Gm = Delta Plate Current / Delta Bias Voltage
Example:
Initial setting
Bias voltage is -25V
Plate current is 40mA
Change Bias voltage to -24V
Plate current is now 50mA
Delta Plate current is 10mA (0.010A)
Delta Bias is 1V
Transconductance: 0.010A / 1V = 10,000 uMhos.
Now repeat, but change the bias voltage in the other direction:
Initial setting
Bias voltage is -25V
Plate current is 40mA
Change Bias voltage to -26V
Plate current is now 31mA
Delta Plate current is 9mA (0.009A)
Delta Bias is 1V
Transconductance: 0.009A / 1V = 9,000 uMhos.
Notice that the transconductance is different for the 1V delta bias voltage, versus the 1V delta bias voltage change in the opposite direction.
That difference in transconductance is what causes 2nd harmonic distortion, and causes 2nd order intermodulation distortion.
Also, Beam Power Tube Transconductance is defined and measured in the same way.
Does anybody need a simple schematic of that test?
So what does the dumm blonde do with those? He runs two in parallel for even more power.
A pair in PSE should get me more power than the OPT's will eat. Two pair in PPP will do about 500 WPC.
Maybe, but a good amp is nearly always running from 20mW to about 2W even on (say my 2 x 80wpc) system.
The only place you know high powers is thru the 25-45hz region with stuff like bass drum transients on uncompressed audio or organs....(very rare).
I have never understood the need for these powers of 70-500W. It makes no sense at all.
In my experience with Scott parts*, (which 7591s were used with), the magic is clearly in the output transformers, while the nice linear class A regions when used with those high gain valves, work out to an excellent sound.
As proof of that D Berning used those transformers* in his EA-230 which the car amp was based on. (actually a Williamson design). That thing to my mind is just way out of date, but they sold a lot.
I do a lot better with much smaller than the 7868, it's a genuine pentode, so vanishingly small IMD, and it's totally reliable with mil spec glass at 220C easy.
cos of chinese spies on forums, most of us have given up letting on what works, cos all they do is go in and buy them all up before you can say "nĭ hăo"!
The only place you know high powers is thru the 25-45hz region with stuff like bass drum transients on uncompressed audio or organs....(very rare).
I have never understood the need for these powers of 70-500W. It makes no sense at all.
In my experience with Scott parts*, (which 7591s were used with), the magic is clearly in the output transformers, while the nice linear class A regions when used with those high gain valves, work out to an excellent sound.
As proof of that D Berning used those transformers* in his EA-230 which the car amp was based on. (actually a Williamson design). That thing to my mind is just way out of date, but they sold a lot.
I do a lot better with much smaller than the 7868, it's a genuine pentode, so vanishingly small IMD, and it's totally reliable with mil spec glass at 220C easy.
cos of chinese spies on forums, most of us have given up letting on what works, cos all they do is go in and buy them all up before you can say "nĭ hăo"!
Last edited:
"500 WPC"
Ear plugs required.
----------------------------------
Umm, that shouldn't be the case unless there is hysteresis somewhere. Probably some digital readout graininess or measurement error there.
Delta I over delta V should be the same for either direction.
2nd harmonic comes from the increasing gm as plate current increases in a SE setup. Usually gm varies roughly as the SQRT of current for power tubes. Until cathode saturation at high current starts to compress that toward constant gm.
Constant gm gives linearity. Class A P-P sums out to almost constant gm, but not quite. Opposing SQRT functions get summed to give a small gain hump in the middle region. That's usually better than class AB crossover distortion at least.
-----------------------
The lack of "butterfly gain plots" (an Audio Precision term) for tube designs has prevented real linear design from ever happening for tube designs historically, other than gobs of NFdbk. Crazy Drive is the only tube design that has solved linearity without NFdbk.
It's a shame, because it is so simple to do butterfly gain plots. An accurate triangle wave generator feeds the amplifier under test. The output of the amplifier gets attenuated and then integrated by an Op Amp circuit and fed to the Vertical channel of an X-Y O'Scope.
The original triangle wave gets fed to the Horizontal channel of the O'Scope. The integrator circuit changes the triangle wave to a square wave (of amplitude proportional to Amp gain conveniently), so one sees a flat line on the O'Scope (scope adjusted to show the top of square wave using an offset and a decent gain setting) Any Amplifier gain variation versus the original triangle wave voltage levels shows up as up/down variation of the flat line on the O'scope. P-P leads to symmetric gain variations that resemble a "butterfly" of sorts.
With this simple test apparatus, P-P crossover distortion/ bias setting, class A P-P gain hump, poor tubes, slew rating limiting, SE 2nd harmonic ... become immediately obvious and easy to decode. Testing an LTP circuit will show the odd harmonic generation, and leads quickly to the solution for that using an inverted Aikido type setup, with a thermionic diode (Rp adjusted triode) for the tail.
This test scheme can also be easily implemented in PC software with a sound card interface, and would be an obvious add-on feature for FFT software.
Ear plugs required.
----------------------------------
Umm, that shouldn't be the case unless there is hysteresis somewhere. Probably some digital readout graininess or measurement error there.
Delta I over delta V should be the same for either direction.
2nd harmonic comes from the increasing gm as plate current increases in a SE setup. Usually gm varies roughly as the SQRT of current for power tubes. Until cathode saturation at high current starts to compress that toward constant gm.
Constant gm gives linearity. Class A P-P sums out to almost constant gm, but not quite. Opposing SQRT functions get summed to give a small gain hump in the middle region. That's usually better than class AB crossover distortion at least.
-----------------------
The lack of "butterfly gain plots" (an Audio Precision term) for tube designs has prevented real linear design from ever happening for tube designs historically, other than gobs of NFdbk. Crazy Drive is the only tube design that has solved linearity without NFdbk.
It's a shame, because it is so simple to do butterfly gain plots. An accurate triangle wave generator feeds the amplifier under test. The output of the amplifier gets attenuated and then integrated by an Op Amp circuit and fed to the Vertical channel of an X-Y O'Scope.
The original triangle wave gets fed to the Horizontal channel of the O'Scope. The integrator circuit changes the triangle wave to a square wave (of amplitude proportional to Amp gain conveniently), so one sees a flat line on the O'Scope (scope adjusted to show the top of square wave using an offset and a decent gain setting) Any Amplifier gain variation versus the original triangle wave voltage levels shows up as up/down variation of the flat line on the O'scope. P-P leads to symmetric gain variations that resemble a "butterfly" of sorts.
With this simple test apparatus, P-P crossover distortion/ bias setting, class A P-P gain hump, poor tubes, slew rating limiting, SE 2nd harmonic ... become immediately obvious and easy to decode. Testing an LTP circuit will show the odd harmonic generation, and leads quickly to the solution for that using an inverted Aikido type setup, with a thermionic diode (Rp adjusted triode) for the tail.
This test scheme can also be easily implemented in PC software with a sound card interface, and would be an obvious add-on feature for FFT software.
Last edited:
Th EL506 is my favourite, it was the first power tube I used in a PP amplifier loosely based on the Lowther LL18.
The EL506 is an improved version of the 7868 but being made in Europe it has the Magnoval base. (Brimar = British Made American Radio - apparently one of the American firms tried to make a dent in the European market and the Brimar tubes are normally of high quality and tight tolerance. E.g. the Brimar ECC81 has a Vhk rating of 250 V which none of the others have. I've got a pair of rare triple mica Brimar E81CC...)
I had forgotten about this EL506 and had locally acquired a NOS quad of RCA 7868 with known provenance. I had asked he seller if he had genuine KT66 but a single NOS GE KT66 was as much as the four 7868 so no contests.
Then late last year I came across a pair of NOS MAZDA EL506 at a give away price (estate sale) - these are relabelled Brimar EL506 because no-one else ever made them.
The data sheet clearly specifies 100K grid resistor for fixed bias and 330K for cathode bias. If you use these values for the current producyion 7591 / 7868 then you will not go wrong but make sure that the filament voltage stays within the specified range as g1 can easily go into thermal runaway.
The EL506 is an improved version of the 7868 but being made in Europe it has the Magnoval base. (Brimar = British Made American Radio - apparently one of the American firms tried to make a dent in the European market and the Brimar tubes are normally of high quality and tight tolerance. E.g. the Brimar ECC81 has a Vhk rating of 250 V which none of the others have. I've got a pair of rare triple mica Brimar E81CC...)
I had forgotten about this EL506 and had locally acquired a NOS quad of RCA 7868 with known provenance. I had asked he seller if he had genuine KT66 but a single NOS GE KT66 was as much as the four 7868 so no contests.
Then late last year I came across a pair of NOS MAZDA EL506 at a give away price (estate sale) - these are relabelled Brimar EL506 because no-one else ever made them.
The data sheet clearly specifies 100K grid resistor for fixed bias and 330K for cathode bias. If you use these values for the current producyion 7591 / 7868 then you will not go wrong but make sure that the filament voltage stays within the specified range as g1 can easily go into thermal runaway.
I have "dis-assembled" a Sylvania 10JA5 and a 12GE5 tube to see what is different between them, since they look alike.
The 10JA5 was of interest since it is a Vertical Sweep tube with curves similar to the 7591 to some extent. The 12GE5 is a Horiz. Sweep tube with efficient lower knee voltages. Vertical Sweep tubes need to be linear for their job. Horiz. sweeps not necessarily, they run as switches.
These two tubes were virtually identical inside except for a few small differences. The heater power of the 10/6JA5 is lower than the 12GE5. 1.0 Amp for 6JA5 versus 1.2 Amp for 6GE5. (0.8 Amp for 7591) And the cathode coating was 1/8 inch shorter for the 10JA5, although the cathode sleeve for both was identical size. The grids were all identical, except the 10JA5 grid1 was very slightly rounded out away from the cathode, while the 12GE5 has a perfectly flat grid1. Cathodes were flat. Plates were identical except the 10JA5 had notches running along near the seams. Grid 2 for both were identical pitch and wire aligned with grid 1. Grid 2 was flat for both tubes.
So apparently the 10/6JA5 does some gm modification across the surface to get the smoother curves it has.
The 10JA5 has a low voltage screen grid versus the 7591, so is not a quick and easy substitute.
It looks like the JJ KT77 may be doing a similar feat for curve modification. So that may be a similar sound substitute for the 7591, if the higher heater power ( 1.4 Amp at 6.3V) can be accommodated. As well as greater height due to being a bigger Wattage tube. And different pin-out.
By the way, the curve modification done to the 10/6JA5 makes for poorer triode curves than the 12GE5. Win some, loose some, trade-offs.
The 10JA5 was of interest since it is a Vertical Sweep tube with curves similar to the 7591 to some extent. The 12GE5 is a Horiz. Sweep tube with efficient lower knee voltages. Vertical Sweep tubes need to be linear for their job. Horiz. sweeps not necessarily, they run as switches.
These two tubes were virtually identical inside except for a few small differences. The heater power of the 10/6JA5 is lower than the 12GE5. 1.0 Amp for 6JA5 versus 1.2 Amp for 6GE5. (0.8 Amp for 7591) And the cathode coating was 1/8 inch shorter for the 10JA5, although the cathode sleeve for both was identical size. The grids were all identical, except the 10JA5 grid1 was very slightly rounded out away from the cathode, while the 12GE5 has a perfectly flat grid1. Cathodes were flat. Plates were identical except the 10JA5 had notches running along near the seams. Grid 2 for both were identical pitch and wire aligned with grid 1. Grid 2 was flat for both tubes.
So apparently the 10/6JA5 does some gm modification across the surface to get the smoother curves it has.
The 10JA5 has a low voltage screen grid versus the 7591, so is not a quick and easy substitute.
It looks like the JJ KT77 may be doing a similar feat for curve modification. So that may be a similar sound substitute for the 7591, if the higher heater power ( 1.4 Amp at 6.3V) can be accommodated. As well as greater height due to being a bigger Wattage tube. And different pin-out.
By the way, the curve modification done to the 10/6JA5 makes for poorer triode curves than the 12GE5. Win some, loose some, trade-offs.
Last edited:
One smart thing about the 10JA5, you set up a stout 40V DC power supply for the heaters (38-42V will do) 1 x 24V winding + 1 x 6.3V winding + rects, and it sets you up a perfect stable multi tapped bias supply in 10.7V steps, 42, 31, 20.5 (CT)....to which you can chain any small pot to set up the -g1 bias.
I notice a number of 60s valve amps used this arrangement with great success.
I notice a number of 60s valve amps used this arrangement with great success.