I have a microphone stereo preamplifier based on ECC82. First stage is feedback and tube rolling has no effect on the response. Second stage is SRPP. Because ECC82 has a high impedance, and the output transformer is not happy with this, I have a bit of bass rolloff. Nothing tragic, about 1 to 1.5db loss at 40Hz. For what i record (acoustic guitar) is not a problem.
The strange thing is that when I test with RMAA, modern JJ tubes (not even the fancy 802/gold) show better graphs than several vintage tubes I tried.
Vintage show at least 0.5db more bass loss, and the distortion is double according to RMAA. However old tubes have a very slightly higher treble response than the JJs. I've tried Amperex buggleboy, Siemens E82CC and Adzam (Philips rebrand) with virtually identical results. Knowing the impedance mismatch, it leads me to believe the old tubes have a higher Ri.
Could this be from construction, or are the tubes worn and not really NOS? Or duds? The Adzam is surely NOS, the seller had dozens of them in sealed boxes.
The second funny thing, one of the channels (fwiw, the one further away from the PSU) has a better bass response (+0.5db) than the other. They are built with the same components, all high quality tight tolerances. A small difference is that I maybe routed the wirings a little bit better being the second channel built (done p2p). Perhaps slight construction differences in the output transformer?
Thanks!
The strange thing is that when I test with RMAA, modern JJ tubes (not even the fancy 802/gold) show better graphs than several vintage tubes I tried.
Vintage show at least 0.5db more bass loss, and the distortion is double according to RMAA. However old tubes have a very slightly higher treble response than the JJs. I've tried Amperex buggleboy, Siemens E82CC and Adzam (Philips rebrand) with virtually identical results. Knowing the impedance mismatch, it leads me to believe the old tubes have a higher Ri.
Could this be from construction, or are the tubes worn and not really NOS? Or duds? The Adzam is surely NOS, the seller had dozens of them in sealed boxes.
The second funny thing, one of the channels (fwiw, the one further away from the PSU) has a better bass response (+0.5db) than the other. They are built with the same components, all high quality tight tolerances. A small difference is that I maybe routed the wirings a little bit better being the second channel built (done p2p). Perhaps slight construction differences in the output transformer?
Thanks!
Yes, you have several variables going on. A schematic of what you're actually measuring and the measurement graphs could be helpful.
In general, the JJ small signal tubes have a different construction than their classic counterparts- in the case of the ECC81, I have also noted significantly lower distortion than other ECC81, so I'm not surprised to hear the same thing about the ECC82.
In general, the JJ small signal tubes have a different construction than their classic counterparts- in the case of the ECC81, I have also noted significantly lower distortion than other ECC81, so I'm not surprised to hear the same thing about the ECC82.
Nowhere near enough information for us to comment, apart from saying that the ECC82 is usually regarded as low impedance (at the anode).
The schematic is here: http://www.gyraf.dk/gy_pd/g9/g9_sch.gif
I have tried ECC88 and ECC99 (with the approapite resistors) in the output and they are much more linear for the taste of the output trafo. There is zero rolloff in the bass (actually the output cap needs to be bumped to some 12uF to push down the resulting resonance) but on the other hand the treble is rolled off a bit.
Attached is a RMAA graph. Big bass is ECC99. Probably tested with 10uF out, needs more. The other is a vintage ECC82. A JJ would have -1db at 50Hz rather than -2. Half the THD. Looks a lot better than the vintage, at least on paper. Also the other channel has 0.5db more bass whatever the tube used.
With the ear it is really very hard to test. Perhaps I could say vintage sounds sweeter. It is not like hifi where you play a CD over an over. Pretty much impossible to play a (string) instrument the same twice. Even if I put matched mics side by side to compare JJ (left) vs vintage (right channel) the mics do not occupy the same exact point in space so differences happen.
I have tried ECC88 and ECC99 (with the approapite resistors) in the output and they are much more linear for the taste of the output trafo. There is zero rolloff in the bass (actually the output cap needs to be bumped to some 12uF to push down the resulting resonance) but on the other hand the treble is rolled off a bit.
Attached is a RMAA graph. Big bass is ECC99. Probably tested with 10uF out, needs more. The other is a vintage ECC82. A JJ would have -1db at 50Hz rather than -2. Half the THD. Looks a lot better than the vintage, at least on paper. Also the other channel has 0.5db more bass whatever the tube used.
With the ear it is really very hard to test. Perhaps I could say vintage sounds sweeter. It is not like hifi where you play a CD over an over. Pretty much impossible to play a (string) instrument the same twice. Even if I put matched mics side by side to compare JJ (left) vs vintage (right channel) the mics do not occupy the same exact point in space so differences happen.
Attachments
ECC82 (100V, 12mA, Ri:6k25) based SRPP has -about- 4k output impedance.
If you drive transformer (parafeed?), the LF rolloff depends of the output impedance, the coupling capacitor and transformer primary inductance.
If you drive transformer (parafeed?), the LF rolloff depends of the output impedance, the coupling capacitor and transformer primary inductance.
To get low distortion from an SRPP you either need very light loading or optimum loading (which usually means a significant loss of gain). I suspect that circuit falls between these two, so is not making best use of the SRPP. Maybe the designer made the mistake of assuming that optimum load is the same as output impedance?
LF rolloff will depend on output impedance, which will depend on valve characteristics. For example, lower mu means lower output impedance.
LF rolloff will depend on output impedance, which will depend on valve characteristics. For example, lower mu means lower output impedance.
Well my main curiosity is why is the JJ different from vintage tubes. But for what my ears and monitoring are worth, I can't hear any significant difference between a channel loaded with say Siemens E82CC and the other JJ. I think i can hear something, but it can very well be the 1cm distance between mics capsules, differences between mics (it is not a close matched pair) or my brain.
I do hear the difference when I use the ECC99 but it's not really better, just different. There is more bass but not as easy and articulate.
In actual use, I find the pre really nice. Very clean (that's what i need) and with the right amount of body, not thin not too fat.
I do hear the difference when I use the ECC99 but it's not really better, just different. There is more bass but not as easy and articulate.
In actual use, I find the pre really nice. Very clean (that's what i need) and with the right amount of body, not thin not too fat.
You would have to measure the JJ to discover what characteristics are different from typical NOS valves. ECC99 is quite different, so different results are to be expected.
Note that if the output coupling capacitor resonates with the transformer inductance than you can expect to get LF results which vary widely with output resistance. Maybe a bigger cap value should be used? However, the inductance seen by the circuit will depend on the external load too so there are lots of variables.
Note that if the output coupling capacitor resonates with the transformer inductance than you can expect to get LF results which vary widely with output resistance. Maybe a bigger cap value should be used? However, the inductance seen by the circuit will depend on the external load too so there are lots of variables.
I did measure the JJ. I don't have a graph now but the response lies somewhere in between vintage 82 and 99 (of course closer to 82 than to 99). THD is exactly half. (99 is even less)
If you refer to measurements like those done on tube testers...maybe I could find someone locally. Perhaps it would be worth to compare spec sheets too.
If you refer to measurements like those done on tube testers...maybe I could find someone locally. Perhaps it would be worth to compare spec sheets too.
>believe the old tubes have a higher Ri.
Measure in-circuit tube current. If they don't bias up to roughly the same current, Gm and Ri will differ. A first-order compensation would be to change the cathode resistors to change current.
If you really want to dig: divorce the tube from the transformer. The tube is flat 0 to 100KHz. The transformer has several response issues with source and load(!) impedance. Looking at half-dB deviations in RMAA may not be illuminating, and does over-emphasize "errors" we never worried about in old +/-3dB days.
Measure in-circuit tube current. If they don't bias up to roughly the same current, Gm and Ri will differ. A first-order compensation would be to change the cathode resistors to change current.
If you really want to dig: divorce the tube from the transformer. The tube is flat 0 to 100KHz. The transformer has several response issues with source and load(!) impedance. Looking at half-dB deviations in RMAA may not be illuminating, and does over-emphasize "errors" we never worried about in old +/-3dB days.
Here is a graph that shows it well. Upper line is right channel, double mid is right with vintage, left with JJ, lower is left with vintage. So there is a way to have the amp well balanced if I mix vintage with JJ 🙂
My best guess now is that the transformers are slightly different.
My best guess now is that the transformers are slightly different.
Attachments
It could be poor transformer matching, but a different bias point would have the same effect. Have you compared DC voltages between the channels? Resistor values?
Resistors are tight, voltages close enough together. From what I have messed with the bias resistors, I need to go pretty wild to budge the graph so noticeably as the mistery difference. What I found out from my bias tests, the designer found a very good compromise between lowering impedance, distorsion and noise.
The resistors are an unusual pair, 1500 up and 470 down. All other SRPP I've seen had equal values. From my tests, more equal values result in more bass cut (with less distortion) or less bass cut but with higher distorsion and a bit of extra noise.
The only thing I couldn't try is to raise the 1500 even more, as I don't have easy access to disconnect it from the tube turret.
The resistors are an unusual pair, 1500 up and 470 down. All other SRPP I've seen had equal values. From my tests, more equal values result in more bass cut (with less distortion) or less bass cut but with higher distorsion and a bit of extra noise.
The only thing I couldn't try is to raise the 1500 even more, as I don't have easy access to disconnect it from the tube turret.
Strictly speaking, if the two cathode resistors are not the same value then it isn't an SRPP but halfway between SRPP and mu-follower.
Maybe the two transformers are from different batches, with different core material? Modern production methods ought to guarantee the same number of turns on each.
Maybe the two transformers are from different batches, with different core material? Modern production methods ought to guarantee the same number of turns on each.
I thought so too, and they are new lundahl, bought together, not cheapos or vintage. Oh well. Maybe I can try to test by jumping wires from one channel to the other's trafo
Do they have serial numbers? Have you asked Lundahl - they might not realise that their iron supplier has poor quality control?
There is such a thing as unit to unit variation and if the rp of the driving tube is towards the high end or beyond the recommended source impedance range for that transformer there will be measurable variations in the frequency response in band.
I use a lot of transformers from a variety of makers in my designs and none are ever identical - in many cases they are very close and should meet the stipulated performance if the specified input conditions are met.
None of this is really surprising, what is the recommended driver source impedance for this transformer?
I use a lot of transformers from a variety of makers in my designs and none are ever identical - in many cases they are very close and should meet the stipulated performance if the specified input conditions are met.
None of this is really surprising, what is the recommended driver source impedance for this transformer?
http://www.lundahl.se/wp-content/uploads/datasheets/5402.pdf
I looked at the data sheet and this transformer is designed for low impedance drive (as low as possible according the data sheet, think op-amps)
Your driver circuit (quasi SRPP) probably is not what they had in mind as a driver when they designed these transformers. Probably insufficient inductance for extended LF response with even a few hundred ohms of source impedance.
This still might be just the right transformer for this application based on the designer's goals.
I looked at the data sheet and this transformer is designed for low impedance drive (as low as possible according the data sheet, think op-amps)
Your driver circuit (quasi SRPP) probably is not what they had in mind as a driver when they designed these transformers. Probably insufficient inductance for extended LF response with even a few hundred ohms of source impedance.
This still might be just the right transformer for this application based on the designer's goals.
Yes it is not what Lundahl made them for, but they sound sweet. The small bass loss is not something I can hear when actually recording. So I am digging into it just for educational purposes.
In any case I pretty much ruled out a funky trafo in my right channel. I did the RMAA tests going unbalanced (skipping the trafo) and there is still a difference between the channels.
Interestingly, it is not so big, meaning the trafo amplifies a bit what is happening in the LF. The trafo also boosts the treble a bit and also adds a solid amount of distorsion.
In any case I pretty much ruled out a funky trafo in my right channel. I did the RMAA tests going unbalanced (skipping the trafo) and there is still a difference between the channels.
Interestingly, it is not so big, meaning the trafo amplifies a bit what is happening in the LF. The trafo also boosts the treble a bit and also adds a solid amount of distorsion.
I find it interesting that the output is coupled from the CCS cathode instead of the output tube plate.
- Status
- Not open for further replies.