Fixed bias is for sure a nice engineer's solution to getting the most out of a design, but nothing beats self-bias for shear reliability and simplicity. My experience is that individual cathode resistors is even better balance-wise, and cross-coupled garter bias is even better still, but garter bias can waste twice the bias voltage, so isn't practical in all instances. Lower volt applications it works fantastic, and cathode bias in general is a nice way to reduce dissipation on older guitar amps that are running lots of voltage across the finals and making you nervous
When building an amp that isn't living with someone likely to pay attention to such things, I use self-bias or garter bias to make it more or less foolproof.
Yup. Knowing this can open your eyes to a lot of things, and also can influence designs... A nice source follower can work wonders in letting finals "push" a little harder on transient higher signal, but the cathodes need to let the extra current through somehow! Combination or fixed bias is the way to go here. Or, better yet, more sensitive speakers or box design.
Also, running of of a better source (like a nicer DAC with better headroom/dynamic range) can make things more apparent due to lack of compression.
Interestingly enough, my 6SN7 push pull is garter biased, and probably does just a smidge over a watt at "full blast" without any audible clipping behavior, and at very high (more than would be comfortable!) listening volumes. I have it hooked up to some DIY karlsonator (with vifa/tymphany TC9FD) speakers that are reasonably sensitive. Hooking this amplifier up to some home theatre speakers will give you much less audible volume for the same signal before sounding questionable...
100V peak vs 1Vrms is 5000x more power...
Today, or during the height of the loudness war, cd recordings regularly peaked out at only 8dB vs rms or even less, but there was also a time where the average recording level was at -18dB, which is a factor of ~63 in power... Now, if you like to listen to music which also has quieter passages, a factor of 100 in power is not too uncommon.
Many different concepts ... any impressions on the sound differences they actually provide ? I guess that there are for sure different effects audible like:
- Really the lowested saturation of the core by matching the currents per PP-Output tube (the explaination given by Menno for the improvement on details with the autobias module) vs.
- Autobias of the current - source setup (which is only really the same current if both tubes behave identically, no ?)
- Classic fixed-bias vs. cathode-resistor
- mixture of the last two concepts ?
- Really the lowested saturation of the core by matching the currents per PP-Output tube (the explaination given by Menno for the improvement on details with the autobias module) vs.
- Autobias of the current - source setup (which is only really the same current if both tubes behave identically, no ?)
- Classic fixed-bias vs. cathode-resistor
- mixture of the last two concepts ?
If implementing "classic fixed bias" : do not stabilize the bias source unless B+ is also
stabilized. Reason is that if both are unstabilized they will follow mains fluctuations
and to large degree be a stable combination ( B+ increases, bias will also increase).
This is both simple and cheap and needs no advanced solutions.
stabilized. Reason is that if both are unstabilized they will follow mains fluctuations
and to large degree be a stable combination ( B+ increases, bias will also increase).
This is both simple and cheap and needs no advanced solutions.
Thanks for the correction, I thought my math was off. Now I understand what you mean. If, for instance, you listen to a classical piece with those extremely quiet parts that you need to turn up to hear, then there is a cresendo and it's shaking the walls. I feel like building a compressor just to deal with that, and movies.
Reducing core saturation at idle is definitely a benefit of autobias, but individual cathode loads or garter bias can accomplish this in an acceptable manner too.
A current mirror on top of the current sink can help the balance, as will individual current sinks. There are ways to allow the current to change with signal level (I think John Broskie called his scheme "sliding bias" in one of his blog writings) but I've not had much experience with them
They can sound different for sure, as cathode bias adds some degeneration to the signal, even when bypassed. Large signal excursion will cause blocking distortion. Cathode bias will have less gain and more even harmonics too... In my opinion fixed bias is one of the best ways to prevent blocking distortion, especially when you run followers to feed the grids. Fixed bias will often sound a bit more clean, especially when pushed hard. It will often measure a bit nicer too.
Depends on how much you skew the control... more fixed bias with around 10-15% of the bias being developed across a cathode resistor can work pretty well and sound great, but unless parallel connecting a bunch of tubes I don't like to use it... you could throw a current mirror under the cathode and run fixed bias, but that needs a little bit of fooling around to make work well, not that it's all that difficult...
Do I sound a little vague, like I'm withholding a bit of info? Sure. Part of that is because I've not found any form of autobias that I truly like yet, and I'm trying to inspire you do dig deep into some older literature and do some research. I find looking at older patents from the sixties and earlier can be quite enlightening and helpful.
Merlin said "power", not "voltage".
So 10V peaks on this "1V signal".
Merlin has engineered recording circuits, and may know something.
"RMS" can be defined many ways. In speech/music a 50mS time-constant is fairly common. On that time-scale, 50ms RMS does tend to be more than a few dB below peak. (I've run many dozens of hours of live recordings, and more than a few processed recordings, through this computation to know if my masters were "loud enough".)
Just implement it like this (attached). Doesn't get much simpler. Metering consists of 10R/0.25W current sense resistors in the cathodes. These should be sized small so they act as fuses of last resort.
The metering is with a 100uA meter with a series resistor to set the full scale. I calibrated it by setting up a 100mA CCS with a power transistor to set the current. Used a 100K pot in series with the meter and adjust to 100uA. Note the series resistance. In this case, it was a little over 7K, and within a couple of ohms of a combination of 12K + 1K. A small on-off-on switch selects each final for bias adjust.
Attachments
Just some info: For a 5mVrms (5cm/s) pickup you can expect rare transient peaks up to 16dB above the 5cm/s level, which is 31.5cm/s. Clicks due to dust and scratches may reach 22dB above the 5cm/s level, which would amount to 89mVpk using that pickup.
May I ask a maybe stupid question here ?
I take a test signal like a sinus, 1000hz, 0db. Its a pcm file coming from a wavegenerator, so a wav file. I will be played by my DAC at 100% volume. My DAC is a Buffalo, so ess9018 with a differential DHT ouput stage btw and gives approx 3V rms or approx. 8V p-p output with this signal.
Sofar I would have thought: Well, this is full digital signal, it will technically not be possible to become louder than this, right ?
So, if we would define the minimum negative grid voltage for the input/driver (in my case only one tube, an e182cc): Half of that 8Vpk-pk plis half a volt safety for no grid current: -4.5V should do it. Right or Wrong ?
Alternatively, If I would add 16db headroom, well 16db would mean six time the headroom, right (2*2*1.5) ? so if I would take this as the guideline: 8v pk-pk times 6 = 48 V envelope. divided by two plus half for grid no grid current would give -24.5V negative bias...which would rule an e182cc out and many other tubes as well. A 801A would be a good input/driver tube than for example....but still: Moving the bias point into the unlinear knee of the triode curves is not what we want either...
So, in this example: which neg. grid voltage would cover your transient requirement for the voltage envelope ? Its understood that we have as well a current requirement for the transient response not to be compromised.
I take a test signal like a sinus, 1000hz, 0db. Its a pcm file coming from a wavegenerator, so a wav file. I will be played by my DAC at 100% volume. My DAC is a Buffalo, so ess9018 with a differential DHT ouput stage btw and gives approx 3V rms or approx. 8V p-p output with this signal.
Sofar I would have thought: Well, this is full digital signal, it will technically not be possible to become louder than this, right ?
So, if we would define the minimum negative grid voltage for the input/driver (in my case only one tube, an e182cc): Half of that 8Vpk-pk plis half a volt safety for no grid current: -4.5V should do it. Right or Wrong ?
Alternatively, If I would add 16db headroom, well 16db would mean six time the headroom, right (2*2*1.5) ? so if I would take this as the guideline: 8v pk-pk times 6 = 48 V envelope. divided by two plus half for grid no grid current would give -24.5V negative bias...which would rule an e182cc out and many other tubes as well. A 801A would be a good input/driver tube than for example....but still: Moving the bias point into the unlinear knee of the triode curves is not what we want either...
So, in this example: which neg. grid voltage would cover your transient requirement for the voltage envelope ? Its understood that we have as well a current requirement for the transient response not to be compromised.
Last edited:
I have once observed 0.5V peaks on a shure M44 pickup using
a 100mhz hp memory scope. But ordinary clicks seems to be in 10-100mV range
Right.
4Vpk is the maximum you will see. If we assume that is a transient peak that is 16dB above 'nominal', which we want to reproduce without clipping, then we are basically saying the maximum 'nominal' audio level at that point in the circuit will be 0.63V peak, or 0.45V rms.
But I'm using numbers for vinyl. Digital signal sources have higher dynamic range than vinyl, so the peaks may be more than 16db! On the other hand the music may be more compressed, so who knows. Just turn it up until you don't like the sound anymore. Congrats, you just found your headroom limit!
Last edited:
Interesting. I don't understand what velocity has to do with level? Could you explain? Thanks Merlinb.
v(t)= dPhi/dt
The faster the magnetic flux changes, the higher the voltage induced in a pick-up coil.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.