In audio tap[e recorderas RF bias is used to work around the magnetic hysteresis of the tape.
But what if we use RF bias in an audio amp with crossover distortion? I tried this without bias: ftp://panteltje.com/pub/rf_bias_against_xover_distortion_no_bias.gif and then this with bias: ftp://panteltje.com/pub/rf_bias_against_xover_distortion_with_bias.gif The blue line is the filtered output at R2/C1
The signal is 1000Hz 8V. The bias, if present, is 2V. The top voltage source is 100kHz 2 V.
Does this make sense ?
The reason I tried this is that I once had a mixer with a switchmode supply, that did feed through to the main amp, everybody liked that sound, but later I found that there was RF from that mixer in the speaker leads....
All sorts of distortions can be blurred out using dithering in ADCs so the answer has to be a qualified yes. The down side is that the amplifier is doing work to amplify this RF. This makes it sound unlikely as a way to save power.
Oh... Thats a new one on me. I have a calculation on this showiing that the the basic reason for distortion reduction is parametric amplification. To wit:
formatting link
If you would care to point out any errors, I might be inclined to change my view. Do you have any papers analysing why you believe HF bias reduces distortion due to hysteresis, and not just the inherent non-linear behaviour of tape?
Probably. I have never really thought about this particular application of HF bias, but the theory is quite general, so it would, on the face of it, appear to explain your pictures.
On a sunny day (Sat, 11 Oct 2008 16:07:05 GMT) it happened "Kevin Aylward" wrote in :
Now that is nice, I first though about writing it out, but then, as I was doing other stuff at the same time, spice was easier to quickly check if the idea worked. In my younger days, before we were let lose in the studios, we got a 9 month in the school benches audio video training from the network (they do not do that anymore I think). One old person, from the radio section, explained the RF bias in tape recorders thing, he explained it very clearly, graphically, in the time domain, with the tape (IIRC) leaving the record head and the magnetisation following the tape's BH curve getting weaker and weaker, and finally settling for one value. Perhaps you can see a similar thing in the spice simulation I presented, if you enlarge the area around the cross-over. Anyways, I will not check your math, you likely are better then me at that, but it seems you were there first :-)
The tape's BH curve _is_ hysteresis, maybe I did not make myself clear?
Wanted to draw ASCII, but this is much better:
formatting link
Look at 'magnetic hysteresis', this is also what you see if you magnetise - demagnetise the tape (iron).
I can think of practical applications, a small RF, and it must be small as it does limit the drive on top and bottom (it should be a pure sine wave not clip), maybe modulated so it is weaker for strong signals?, and a setpoint of the power amp just below cutoff. The extra RF would use little power, perhaps less then a little bit of bias would, as it would see a high impedance anyways (speaker) or even a wide LC filter?
Yes, and you can fly to the moon by pulling up on your bootstraps.
What you're doing is dithering the signal around zero, hoping to get at least one transistor into a bit of its active region. Which is not unlike trying to push-start your car by throwing rocks at it.
In the real world you're going to have problems:
(1) The transistors are not going to be 2N3904's, but power transistors or FETs, usually not designed for RF. You didn't simulate this.
(2) You may improve the crossover region, but you're going to mess up the signal peaks. You didn't simulate this.
(3) Unless your customers have a FCC license, you'll need to add a passive low-pass filter to the output, and maybe some shielding.
(4) When your cost accountant sees the extra cost of the HF transistors, the filter, and shielding, they'll fill their pants.
So, if I understand it correctly, if you add RF wiggles to a nasty O/P stage you reduce the distortion.
I have a feeling that that you only need to add RF when the signal is approaching the crossover zone, and not when the input signal is significant. Is this a fair assumption?
On a sunny day (Sat, 11 Oct 2008 22:59:47 +0200) it happened Martin Griffith wrote in : So, if I understand it correctly, if you add RF wiggles to a nasty O/P
Yes, but in doing so you would [have to] modulate the RF signal in a special way. Because the signal would be minimum for both the pos and neg audio peaks, while maximum for zero....
One can ask how much you can hear crossover if the amplitude is maximum....
A silly question, since I haven't stuck a scope across an audio amplifier for at least 10 years, and I really suck at high power audio (it really,really makes me fall asleep)
How large is the crossover zone, a couple of diodes worth of V, 1400mV or so, into 4 ohms?
On a sunny day (Sat, 11 Oct 2008 23:25:51 +0200) it happened Martin Griffith wrote in :
I think, from a distortion POV, say that area where a change in input voltage does not result in a [similar] change of output voltage. That is why I used the NPN-PNP pair without bias, so 1.4V say. But of course, if we were to do this for thermal stabilisation, you could bias it so it was only .2V say. If we were to do this to improve further on Yeore's .000xxx % distortion numbers, then maybe we should refer to a 'non-linear' zone, and that could be any size, but likely small. The smaller that size the less RF we need, so I have been thinking about making a separate box that just adds some millivolts of RF to the input, and a separate loudspeaker filter. The both could be sold for 10000$ to audiophiles, of course their amp would have to have the required bandwidth, say 200kHz, but I think most real audiophiles have that sort of amp. ...
The BH curve shows hysteresis, but that aspect is not important as far as the basic1st order distortion is concerned. For the analysis it is assumed that the loop area of the B-H curve is zero. I don't see how HF bias could reduce the *additional* distortion of hysteresis.
On a sunny day (Sun, 12 Oct 2008 08:56:38 GMT) it happened "Kevin Aylward" wrote in :
But it does, I think your model needs some refinement perhaps? Although your model porves a general case for reducing distrotion by using RF addition, it is perhaps a bit too simplified for the magnetic tape case.
I could be wrong, I have forgotten most of that tape stuff (40 years ago huh), what do you think?
How do you know that? Have you done, or know of actual tests that distinguish between non-linearity and hysteresis effects?
Its a more complicated to include memory (phase) effects, one needs a voltear seriers, or such like, but even this I think is a problem for hysteresis.
I don't think so, as far as basic features.
It accounts for some main, known factors. For example, the nyquist limit is
2 X fmax, this analysis shows that you need at least 3 times fmax for the bias signal, and this is exactly what is done. Typically bias frequencies are 60khz-80khz.
Kevin Aylward snipped-for-privacy@kevinaylward.co.uk
On a sunny day (Sun, 12 Oct 2008 14:23:37 GMT) it happened "Kevin Aylward" wrote in :
Ah, what I tried to say is that any distortion is reduced, also that caused by hysteresis. This because audio tape recordings made with bias are pretty much clean. The required bias amplitude depends on the tape characteristics (BH curve among other things). So I would expect these parameters to be present in your math, if it represented a compete detailed model. Maybe this site can be of some help:
formatting link
Look under 'V', where its says: 'Setting the peak bias field approximately equal to the coercivity creates a curve which is more linear (at least for low values of H!)'
Reminds me of the 'Mule' amplifiers I used to build in the 70's, output stage was a Ge type PNP (forgot which) + Si type NPN (2N3055). Biasing was just a 1N4004, thermally coupled to the output stage. No extra R in series, no trimming. Idle current about 50...100mA. Sounded very good for its purpose: guitar amplifiers. Did a whopping 20W 'rms' (O.K., with a bit of clipping) with 4 transistors.
On a sunny day (Mon, 13 Oct 2008 00:12:52 -0700 (PDT)) it happened snipped-for-privacy@sushi.com wrote in :
You are right, much of the benefit is lost if I add an inductor in series with the load: ftp://panteltje.com/pub/l_without_rf.gif ftp://panteltje.com/pub/l_with_rf.gif
So the dead zone should still be a small as possible.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.