Thanks John. I guess this one below fooled me. I will try a small series resistor again with the lamp. I felt the NE5532 should get to - 90dB. I think it had enough current.
I was old germaniums can short circuit in storage. They seem to grow crystals to the wires.
Solutions - Injection-Lock a Wien-Bridge Oscillator
I was old germaniums can short circuit in storage. They seem to grow crystals to the wires.
Solutions - Injection-Lock a Wien-Bridge Oscillator
If interested in lamps for oscillators this is eactly the lamp I chose ( by coincidence ). Seems 430R was the resistor of choice like mine. Many lamps are 40 mA 28V, 700R hot 60R cold. At my 2Vrms it looks to have 215R which is about where his graphs are. What I now remember is the lamp is best worked this way. Ideally to close to the cold resistance , perhaps 4 times that.
Lamps for stabilizing oscillators
Lamps for stabilizing oscillators
Nigel- I think most lamp stabilised oscillators will work OK at 5-10kHz. This is fast enough for the lamp filament to be almost unable to follow the signal. I should have said that the distortion in my circuit reached a lowest point between 1 and 10kHz.
The TT version gave a much lower distortion at low frequencies. At 1Hz you could see the filament brightening and dimming in time with the sinewave yet still giving a low distortion (perhaps a little above 0.1% if I recall - maybe I'll dig out the oscillators some time and measure them again) but would be several % otherwise.
I generally reckon that you need about 1V across the filament to make any difference. Mainly because of the square law (power~V^2) that anything below 1V rapidly diminshies the applied power so does not change the temperature much.
I've used mains (240V/15W) lamps too. They have bigger thermal mass than the small torch bulbs, and give better LF performance in Wien bridges but may need nearer 2V, so requiring at least 4V output (6V without splitting the feedback path).
My point is that if you provide most of the feedback (say to give a gain of 3.2-3.3 (just enough to start)) with resistors and "dilute" the feedback from the filament lamp, even though the lamp has to then run at a higher % of the output voltage the "dilution" reduces the distortion it introduces.
I think your reference is fine - it's a fixed frequency of 5kHz!
The TT version gave a much lower distortion at low frequencies. At 1Hz you could see the filament brightening and dimming in time with the sinewave yet still giving a low distortion (perhaps a little above 0.1% if I recall - maybe I'll dig out the oscillators some time and measure them again) but would be several % otherwise.
I generally reckon that you need about 1V across the filament to make any difference. Mainly because of the square law (power~V^2) that anything below 1V rapidly diminshies the applied power so does not change the temperature much.
I've used mains (240V/15W) lamps too. They have bigger thermal mass than the small torch bulbs, and give better LF performance in Wien bridges but may need nearer 2V, so requiring at least 4V output (6V without splitting the feedback path).
My point is that if you provide most of the feedback (say to give a gain of 3.2-3.3 (just enough to start)) with resistors and "dilute" the feedback from the filament lamp, even though the lamp has to then run at a higher % of the output voltage the "dilution" reduces the distortion it introduces.
I think your reference is fine - it's a fixed frequency of 5kHz!
Last edited:
John. Good to have your input. If I can reduce the > 40 kHz harmonics I will be happy. Rosen quoted needing distortion - 20 dB better than what is measured to be able to say a test result holds up. Doing a crude estimate as Rosen had to he said a 16 bit signal is 16 x 20 x 0.30103 dB = 96.33 dB. That means -117 dB is where he needed to be. If I remember he did it using NE5534 so it gave me hope ( State Variable Filter type ). 0.01% THD is - 80 dB. Thus - 100 dB would be fine. As my test bed JLH was giving about 0.025% at 6 watts 1 kHz I have to guess my old RA53 Wien is in the ball park. I will carry on. 1 and 10 kHz are ideal .
All the time I have been at DIY Audio I have sensed a reluctance to measure things. I would say that sometimes what gets built is not correct. My first JLH had a highly damped oscillation at circa 1MHz. 1 minute to find and 1 hour to fix. It was subtle and could have been mistaken as the JLH sound. There was clues in the noise floor. The sine wave tests made it obvious.
All the time I have been at DIY Audio I have sensed a reluctance to measure things. I would say that sometimes what gets built is not correct. My first JLH had a highly damped oscillation at circa 1MHz. 1 minute to find and 1 hour to fix. It was subtle and could have been mistaken as the JLH sound. There was clues in the noise floor. The sine wave tests made it obvious.
YouTube
Have found some ASZ17 in my stock;-)
Sounds great. More blackness than the 2N268. Higher tempo, but a little bit KRZZ in the sound - material is a cast, brittle - any correlation;-? Little bit more 3D. More joy of playing. Not as clean as the BD318. But great.
Could replace the wirewound resistors with film. Should counter the krzzz;-)
The most in this diy-audio-forum love the sound of graves-of-parts - I have read;-)
Have found some ASZ17 in my stock;-)
Sounds great. More blackness than the 2N268. Higher tempo, but a little bit KRZZ in the sound - material is a cast, brittle - any correlation;-? Little bit more 3D. More joy of playing. Not as clean as the BD318. But great.
Could replace the wirewound resistors with film. Should counter the krzzz;-)
The most in this diy-audio-forum love the sound of graves-of-parts - I have read;-)
John, I've done a few tests of resistor damped lamp. 120R + 40 mA 28V Lamp lower feedback arm. 430R upper feedback arm. Output at 1.30 V 1 kHz and 2.6 Vrms high output. There might be a marginal improvement. The sine wave has two phases of stability now. One initial of about two seconds and then a lower output second phase. It looks a bit nicer on the scope. As yet distortion about the same as before. One has to guess the lamp plus resistor is about 215R. That's about 35R above the lamp when cold. That looks about as good as it can be. It also keeps the two op amps nicely set at 6 mA. The positive feedback loops are 16K + 1nF or 10nF. Your point you make about the speed of the filement is very likely the real limititation. I might build a cheap Chebishev output filter.
Try carbon composition resistors ( Bakelite plus graphite it is said ). They are available as used with pulse lasars these days where they are more reliable. They have a darker sound much like 1950's amplifiers. They are non inductive as a bonus. Not as good as metal foil types like Tyco if you have deep pockets.
Some smaller germanium transistors are quite fast and give moderate current. I could imagine they could do a good job. Usually the PNP's do a better job.
I did not see your YouTube links, good good good !!
I listen every time this song to each modification of one of my jlh
YouTube
if you want to know if your bass goes down really well
i have no doubt about it
if you have two minutes off peace,listen to this playlist .
i have broken two ampli with her
YouTube
I went to UK company Cricklewood Electronics who have plenty of germaniums. What they have is expensive and mostly I would guess for repairs. What I did notice was high grade PNP power devices that on paper are not too far from the NPN's used by JLH. The only big doubt is ft at 450 kHz. It's enough if class A. Current and gain are not too bad. Germanium's are not safe if they get too hot so no puny heat sinks. I would guess a design using BC327-40 as driver could still return less than 0.1% THD at 20 kHz at 30V regulated and meet the 10 watts spec into 8 ohms. It might even have a respectable power bandwidth. I measured a Hypex at 47 kHz - 3 dB if I remember correctly that due to how it works sounded bright and mostly very good. I suspect a germainium PNP verion could better that. If anyone is interested it's own switch mode power supply made it sound brighter ( it should due to residuals ). I had a tough job making a linear supply that had it's punch. To measure it was very hard. I looked up AP and although very different in how I did it got a way that worked. It was 12 TL074 sections ( 3 x TL074 ) as active inverted unity gain first order filters at about 186 kHz ( - 3dB ). Surprisingly although far too close to TL074's limits it was OK which proves the inverting designs work. Even the distortion was OK. 8 section good enough. Non inverting or bipolar op amps didn't work very well. I suspect much of the said better sound of JFET's is simply how they react with RF. Bipolars cause rectification which shifts the op amp working points.
The only problem here is a PNP germanium could be ideal for a well designed class AB or feedforward design. Analysis of the Quad 405 showed ft not to be very important in that design, 450 kHz would be OK as the correction comes from the fast class A stage. Where the 405 is flawed is the distortion cancelling is just a way of building a cheap OK amplifier, it was really able to be better than that. Interestingly Vce of germanium is about the same as Vbe, the low Vbe suits feedforward ( Audio FET like BUZ900 even better due to soft turn on and 3 mA current at 0V bias ). PNP T03 at £4 a piece would be fun. The JLH might just be where they could work. A bit like using Crossply tyres instead of Radials, mostly for nostalgia ( Michelin XAS being the 2N3055H of tyres ). I found this device as a best buy. The company has others. Always check as one was a silicon wrongly listed.
Semiconductors | Transistors | Diodes | Thryristors | Triacs| ICs | Cricklewood Electronics
AD143 datasheet,datasheets manu Page:1==TRANSISTOR | BJT | PNP | 40V V(BR)CEO | 10A I(C) | TO-3
==[未知厂家] pdf datenblatt - - WWW.ICPDF.COM
The only problem here is a PNP germanium could be ideal for a well designed class AB or feedforward design. Analysis of the Quad 405 showed ft not to be very important in that design, 450 kHz would be OK as the correction comes from the fast class A stage. Where the 405 is flawed is the distortion cancelling is just a way of building a cheap OK amplifier, it was really able to be better than that. Interestingly Vce of germanium is about the same as Vbe, the low Vbe suits feedforward ( Audio FET like BUZ900 even better due to soft turn on and 3 mA current at 0V bias ). PNP T03 at £4 a piece would be fun. The JLH might just be where they could work. A bit like using Crossply tyres instead of Radials, mostly for nostalgia ( Michelin XAS being the 2N3055H of tyres ). I found this device as a best buy. The company has others. Always check as one was a silicon wrongly listed.
Semiconductors | Transistors | Diodes | Thryristors | Triacs| ICs | Cricklewood Electronics
AD143 datasheet,datasheets manu Page:1==TRANSISTOR | BJT | PNP | 40V V(BR)CEO | 10A I(C) | TO-3
==[未知厂家] pdf datenblatt - - WWW.ICPDF.COM
The AD143 was about the best I could find if new old stock verses price. It looked about like the silicon JLH used except they were 4 MHz. 10A 35V gain 30-200 @1 amp is not bad. If less good than that I wouldn't bother. Even so if they sound better than a 2N3055 it could just be collector to base capacitance which is implied in the 450kHz ft. It's very sad to think that germanium is better when a 10 pence 1nF capacitor added to 2N3055 better still ?I did measure some marginal stability issues with my JLH. I wouldn't mind betting all have that? If you do make it stable it will improve the sound. Don't go too far on that. An oscilloscope is required. My cheap JLH oscillator also. It's good enough to give reliable impressions of your work, make sure you test in and out as the JLH might have less HF distortion! I saw it as a + 6db increase in the noise floor between 500 kHz and 3 MHz on my version of the JLH. I fitted a 33pF to my BC337-40 driver and made it go away. I dare say 22pF would have worked if I had some.
Last edited:
Picture: What we are talking about.
The MJ2955! sound a little little bit lazy and stately/solemn, more slowly, compared with BD318 and ASZ17. Fine, great stage, minimal more flat, but there is a twisted "krzlzz" in the sound, finer quieter than the "KRZZ" of the ASZ17;-)-; A little bit more strenuous. Not as contoured and focussed as the both mentioned. The "bass" a little bit bloated, with echo.
The MJ2955! sound a little little bit lazy and stately/solemn, more slowly, compared with BD318 and ASZ17. Fine, great stage, minimal more flat, but there is a twisted "krzlzz" in the sound, finer quieter than the "KRZZ" of the ASZ17;-)-; A little bit more strenuous. Not as contoured and focussed as the both mentioned. The "bass" a little bit bloated, with echo.
Attachments
![IMG_20180118_171545[1].jpg](/community/data/attachments/593/593546-fc58dfffa1729ed301544246aeb25d03.jpg)
![IMG_20180118_191703[1].jpg](/community/data/attachments/593/593921-cfac4f128f88c7c28bc5104fb9ac9979.jpg)
My suggesstion, again and again: case as small as possible: definition of current. Per less diameter, form and less material: "sound". The most TO-3 are made of questionable materials.
fT is not necessary for sound. May be, the less MHz, the better;-?!?! All other parameters "for the cat"-)
Sound is not a resultat of the common specifications.
fT is not necessary for sound. May be, the less MHz, the better;-?!?! All other parameters "for the cat"-)
Sound is not a resultat of the common specifications.
I would speculate that you are not listening to the BD236 at all. What you are listening to is how easy it is to drive. As BD236 mostly is a nothing special device, it's higher gain might be what you like. When a valve like 2A3 or even EL34 one can say it has a sound. A transistor of 3 MHz ft with a typical gain of 100 has no real sound in a high feedback design like this. However like the ideal chain-cog on a bicycle it can be the right thing you need. If you sincerely hear the BD236 as good I suspect TIP36 would be very similar. If you play with the gain of the amplifier you should notice a larger change in sound. Higher gain although more distorted can sound more open. If inside sensible limits I would not say it's a like for distortion. Play with it and see what you get. Very often 400 mV is a good sensetivity.
I plan to make a preamp that can drive headphones of 32R. If so one can drive the inverting inputs 220R . That should sound better. Don't try to drive it with a normal preamp, it won't work.