Audio Bandwidth

[K9ACT]

I understand the subjectivity of all this but in spite of the fact that most of the people I talk to say my audio is very good, there are two gentlemen out there who have issues with the overall frequency response.  I respect their opinions and took the advice of one of them and measured the frequency response of my speech amp and came up with some numbers but I do not know how to interpret them.  I took the advice of the other and bought a graphic equalizer but that is another story.

I have not named them out of courtesy but if they wish to be known, feel free.

Anyway,  I put an audio generator at the mic input and a scope and voltmeter at the output of the transformer driving grids of the 813's and then plotted these numbers on graph paper.

What I came up with looks exactly like an airfoil, i.e.,  the cross section of a wing.

It climbs rapidly from 20 hz to 1khz where it levels off and very slowly falls off to 10kc where it is about the same level as 200HZ.

I am not sure how to relate this to DB's but I ploted everything as linear functions and the data is in mv.

A few of the numbers are:

20......... .003
75             .030
200            .095
500            .200
1000           .275
2000           .316
4000           .285
6000           .200
7500           .175
10000          .113

So, the question is, how do I relate this to frequency response and what is a typical good frequency response of an am transmitter?

My project is at the link below but I have now replaced the Ranger with a homebrew speech amp and Kenwood 460.

http://schmidling.com/radio.htm

Thanks,

Jack K9ACT

[w3jn]

voltage ratio (in dB) = 20 log v1/v2

Let's say 0 db reference is .28V (1000 and 4000 hz points)

you're 20 dB down at 75 hz, 10 dB down at 200 Hz

8 dB down at 10 KHz.

I would say you have the telephone company beat, but not by much 

[Steve - WB3HUZ]

The response in dB can be found by 20log(Volts Input)/(Volts Output).

So for 20 Hertz you have 20log(.003/.275) = -39 dB referenced to 1 kHz.

[K9ACT]

Thanks for filling that hole in my data base.

If I understand and did the math correctly, I come up with "flat" (as in 3 db) from 500 to 6500 hz.

If this is a little better than a telephone, that is one useful reference point but I am not sure what would be considered "good audio", whatever that means.  Broadcast quality aside, what would be a good goal to aim for?

For my own edification, I looked up the freq response of the Ranger and it says 3 db from 250 to 3000.  It would have more low and less high than mine.

The real hooker though is that the criticism of mine was that it is lacking in highs not lows.  This would lead one to think that it is the modulator and not the speech amp but if I am stating with a "telephone" it's not at all clear how to pursue this.

js


 

[w3jn]

Use a scope to look at the RF output of the transmitter (DO NOT connect the scope directly to the RF output, use a clip lead as a halfassed loop antenna).  Then modulate with the tones again and look at the mod % of the envelope.  You should be able to get a good idea of the response of the whole system this way.

Beware that CW tones into the modulator create severe service indeed.  Best to test for a minute, off for 3 or 4 to let things cool down.

[Ian VK3KRI]

Thanks for filling that hole in my data base.

If I understand and did the math correctly, I come up with "flat" (as in 3 db) from 500 to 6500 hz.

If this is a little better than a telephone, that is one useful reference point but I am not sure what would be considered "good audio", whatever that means.  Broadcast quality aside, what would be a good goal to aim for?

Well my aim is for audio that jumps out at you when you tune across it , but sounds natural and pleasant.  Achieving both at the same time is probably beyond my capabilites, but its the holy grail to which I aspire...

Your goal should be what you want. Is it unsolicited reports of 'Broadcast qualtiy audio OM'  or 'Your just moving the meter but I can copy you 100% thru the static crashes'  or both at different times? 

For my own edification, I looked up the freq response of the Ranger and it says 3 db from 250 to 3000.  It would have more low and less high than mine.

The real hooker though is that the criticism of mine was that it is lacking in highs not lows.  This would lead one to think that it is the modulator and not the speech amp but if I am stating with a "telephone" it's not at all clear how to pursue this.

js

The communications path consists of of more than the transmitter. At least 1/2 the path is out of your control. The receiver at the far end, the speaker/headphones at the far end , and worst of all, the listener at the far end  may do very nasty things to your audio.  All you can do here, is tailor the signal to what you would expect a reasonable receiving station to be (and make sure your own receiving setup is better than that!)

You have more control over the transmitting part, you, microphone, audio chain, transmitter. Unfortunatly (well at least for me)  you're stuck with your own voice, although some vocal training could probably do wonders for the average hams voice, I don't think I've ever seen that mentioned as a way of improving the sound of your station. 

Ok having removed what is beyond our control we are left with mic, audio chain, transmitter.  Personally  I like to have all the frequency and dynamics control in the audio chain, rather than the transmitter, but  thats not always possible.   

Record some audio through your mike and normal preamp. Does it sound distorted or have execcive peaks or troughs in the frequency response? This is not always easy to tell. We get very used to 'crunchy' audio, particuarly speech. Some sort of immediate A/B comparison will give you a better guide. Your transmitter can't sound good if the signal going into it is already bad.

My next step would be to check the transmitter itself. Pick up some RF from the output and couple to a dectector that you can feed into an audio amp(I'm sure I've seen some pickup designs posted here before) . Modulate with a lo distortion sine wave and listen for distortion. Compare with the same sine wave straight into the amplifier. Tweak whatever is available for least distortion. (1K is probably a reasonable frequency to use. A Class B modulator will increase distortion at lower frequencies.)

More negative feedback will decrease distortion due to  the bits inside the loop and flatten the frequency response a bit.

Assuming your audio input is now clean, and the transmitter has as little distortion as you can get, now start     
tweaking eq and any dynamics control (clippers, compressors etc). Record audio off your own receiver with the RF gain control down far as possible.  Record yourself in QSO , complete with mumbles, leaning back in chair etc . Saying 'test 123' into the mic is not a real test.  If you can hear a compressor working, its wound up too far.

Don't keep tweaking with every bit of feedback you get about too much or not enough bottom or top end. There is always going to be someone whose definition of 'best sound'  doesn't match yours - and ultimately your the one who has to make the call about how your station sounds.

Anyway, thats my advice, I'm sure there will be plenty more to follow...

                                                                              Ian VK3KRI

                                                                                    Ian VK3KRI

[Steve - WB3HUZ]

You will sound pretty good with the current high end, if you can improve the low end response down to about 100 Hz. If you get response down to about 50 Hz, you will sound as good as anyone else on the air.

[WBear2GCR]

Below 500 to 200 Hz you're about 4 dB down... now an octave below 500 is 250Hz., so we want to know how you look at say 125Hz. What I see is an approximately 6dB/oct roll off below some frequency around 500Hz. or so... plot the response on a log graph to see it better.

What that means in practice is that you can correct that part of the response very simply within the modulator with a properly placed single cap or inductor (and perhaps resistor).

[ ...What you also need to know next is where/if the transformer/circuit starts to fall off like a rock going down to "20Hz." (or does it does fall off more quickly?). ]


In terms of a circuit to correct this part, you'd pick a cap that when "shunted" to ground off the signal path in the speech amp would start to roll off at some frequency around where the circuit now starts to roll off moire than at the 6dB/oct rate. Now you insert the shunting cap circuit to ground, but add a resistor in series to ground that "shelves" the response drop off at the ~500Hz. point. The resulting curve is now roughly FLAT below 500 Hz. down to the point where the LF dropped like a stone.

Ascii art:

0db>-------\
                  \
                   \    <---6dB/oct slope from shunting cap with series resistor "shelf"
-12db>           \_________

20hz        100
                     500          10k

--- the resulting response with the transformer rolling off below 500Hz. is nominally flat. ---


A few tests and iterations of caps and resistor values and you will arrive at the best compromise.

Next step is to determine what is causing the extra fast drop down at some LF freq. Is it the transformer, or is it the coupling caps in the speech amp and mod circuit?? Or perhaps an interstage transformer?

(Checking the response of the circuit up to the grids of the modulator tubes is a very good idea, to make sure that this signal is flat in the first place. Before doing anything else, that's where I'd start. IF it is flat, then solutions like the one I suggested above will be in order... that dovetails with jn's & Ian's suggestion of checking the actual transmited envelope)

You can equalize it externally, OR you can continue to do the EQ inside the speech amp/mod circuit.

You have two choices here - you can ignore the LF drop below ~125Hz. or so, or equalize it.


The initial circuit that I suggested will have two effects - first it will pull the overall gain of the speech amp down about 12dB from where it is now, so you may have to build in a little extra gain somewhere if you a short on gain (use higher gain tubes, etc...), and secondly it will tend to flatten out the LF drop a bit anyhow - meaning that the extra LF EQ required may be less or unecessary when you listen to it.

The other way to go is with an external EQ.
For that, the graphic EQ is one possible way to go. However, keep in mind that it is usually best to DROP gains than to boost! Boosting tends to create overloads, cutting will not. The only thing the transmitter cares about is the resulting response, not how it was created, assuming nothing is being clipped! That's why I prefer cutting over boosting. A little bit of boost may be necessary, but if you find a slider up at +12dB, you can rest assured that you have zero headroom inside that silly solid state EQ... not a good idea.

Alternative treatment 2a. is to wrap feedback around the mod iron, but I do not like that in most cases for assorted reasons, even though HLR does...     (I suppose if you roll off the FB at freqs above ~10kHz., you might be safe(r) )

For the HF drop, you probably would do well to add the EQ (if any) that is needed externally, or very judiciously inside the speech amp. Since most simple HF EQ that is added inside a circuit will continue to boost more as you go higher in freq, that can be a problem as it may add nasties that you'd be better off without.  Essentially flat at 6k seems like it is sufficient, since that is a 12kHz. wide signal... dorking with the response within the 2 to 6k region will only need slight cuts and boosts on an EQ, assuming it has filter widths that are tight enough and freq centers that do you some good...

Hope these ideas help...

             _-_-WBear2GCR


Oh, keep in mind that for most iron applications, as the power level goes up so does the saturation, then the power handling at the freq extremes evaporates and the response goes away... so checking at very low power levels will likely yield best-case LF response results which may or may not mirror real operating conditions.

[WBear2GCR]

Which reminds me... the secondary of the mod iron should be *loaded* when testing freq response. That means in practice, if the RF section is disabled, and B+ is not being sent to the RF finals, that you need a load resistor in order to see what the response of the iron really is!

An unloaded secondary is not the real response of the iron at all.

Although I have not done it, I'd expect that if the B+ is applied to the secondary, then the load for the secondary should be a large power resistor capacitively coupled to the secondary to keep the B+ from running to ground. The power resistor being of the order of the impedance that the mod iron is expected to see... Although if that is high enough, I suppose the B+ might possibly be applied to it as well... Hmmmm.... 

But you do not need the B+ applied to the secondary (and why would you?), the only diff will be in the very LF response, where the saturation of the secondary due to the B+ being applied will have an effect. So I guess skip the B+ and dump a Kohm resistor of suitable power and value across the secondary to see the real response of the iron...



                               _-_-WBear2GCR

[k4kyv]

Jack, here is my audio data that we discussed over the air the other night.


This is my pre-emphasis curve:

Let 200~ be the reference point @ 0 dB.  Response is flat down to the lower limit of the audio chain frequency response, approximately 40 Hz.

Frequency      Level, dB
200~                   0
600~                 +1
900~                 +2
1060~               +3
1325~               +4
1600~               +5
1900~               +6
2325~               +7
2850~               +8
3400~               +8.7
3800~               +9

From 3800 Hz, response is flat to the upper limit of the audio chain, beyond 11,000 Hz.  Above 11,000~ peaks and dips begin to show in the response curve, although response extends beyond 15,000~

Under congested band conditions, I use a passive low pass audio filter with extremely sharp cutoff beyond 3400~.

Under more normal conditions, I use a more gradual passive low pass audio filter that begins to roll off at approximately 5000~, with essentially complete attenuation beyond 7500~.

Here is some information on pre-emphasis curves for voice transmission.  I use less restricted frequency response than George suggests, but I have found that his idea of the rising upper midrange response curve works very well for my voice on my transmitter.

Enhanced AM by George Bonadio, W2WLR

[N3DRB The Derb]

neither here nor there, Don has got one of the oldest D-104's I've ever seen. It's got the LARGE metal I.D.tag from when Astatic was in Youngstown. I had one not quite as old, with the smaller metal tag from same.

[K9ACT]

Here is some new info since I talked to Don last.

I have not gone beyond the speech amp because it gets more complicated and I want to work out all the easier issues in case I don't have to go there.

I suspect the EQ (which is on the way) will solve the high end problems and I will try to duplicate Don's numbers for starters.

I stumbled on a significant improvement on the low end.  Actually, it was suggested by Dan Hopper K9WEK and seems to have helped a lot.

I don't recall exactly where I was at last time we talked but all that low freq roll off is in the interstage transformer.  If I measure before it, it looks pretty good.

I ohmed out the interstage transformer and drew up a schematic and hooked it up so that all the wire is used, both primary and secondary.  Just ignored the fact that the impedence is only about 4000 ohms.

I don't have the numbers in here but the 3 db point went from 500 hz to less than 200 and there is lots of signal left at 20 where there was none before.  I was hoping to catch one of you on the air to see if you could tell any difference. 

I need to run the test again to verify the numbers as I could have done something dumb.

Thanks for the help,

Jack K9ACT

[WBear2GCR]

More turns on the primary = more inductance = lower F3 frequency.

That's at least part of it.

You haven't said if you are running P-P before the interstage or single ended.
If single ended, then if the interstage is NOT gapped, or maybe even if it is, you might really want to consider "parafeed" which is about the same as Heising but used in audio circles to get the DC off the primary of a SE circuit that drives a transformer.

The parafeed trick tends to also improve the HF response... 

If the transformer is NOT gapped, you might consider taking it apart and setting the laminations up with a gap... that will also improve the performance on the low end...

             _-_-WBear2GCR

[K9ACT]

More turns on the primary = more inductance = lower F3 frequency.


You haven't said if you are running P-P before the interstage or single ended.

It is SE.

If single ended, then if the interstage is NOT gapped, or maybe even if it is, you might really want to consider "parafeed" which is about the same as Heising but used in audio circles to get the DC off the primary of a SE circuit that drives a transformer.

Don't know what parafeed is I have a 500 mf cap between the speech amp and the transformer for that purpose.  Does that count?  BTW, the speech amp is a 20 W solid state module.

How do I know if the transformer is gapped?

As a point of interest, I got the EQ today and set it up and it seems to work great until I turn on the HV to the RF/Mod decks.  It sets up a squeel that I can not get rid of.  As soon as I turn up the audio enough to start modultion, the squeel starts.

Any ideas?

js








The parafeed trick tends to also improve the HF response...  ;D

If the transformer is NOT gapped, you might consider taking it apart and setting the laminations up with a gap... that will also improve the performance on the low end...

             _-_-WBear2GCR

[WBear2GCR]

Ok, but I don't know what you've got going on there... so at this point it's shooting ducks in the dark?

Perhaps a block diagram or the actual schematic posted would be best?

A single ended speech amp that feeds an interstage tranny through a 500mfd cap?
Seems a bit too large, unless the input Z to the tranny is very very low indeed.

What exactly is this speech amp? Homebrew? A commercial amp?
It has DC at the output? What power is it? What is on the collector of the output transistor(s)?? etc...

The squeal is feedback or a parasitic.
Or RF getting into the silly EQ... probable.

A gapped transformer does not have the laminations, which are likely E & I shaped, interleaved. Meaning they go in opposite directions on alternating laminations. Otoh, if you put all the "E" lams together, and then all the "I" lams together, the I would be like a cap on the end of the E... that flat spot is the gap. Usually it is adjusted for appropriate magnetic characteristics with a thin spacer. This is also how a "swinging choke" is constructed, fyi.

If you have a cap in series with the primary, you don't need to worry about DC or the gap or lack thereof.

Suggest you google for "parafeed" ... a good read.

                _-_-WBear2GCR

[k4kyv]

A gapped transformer does not have the laminations, which are likely E & I shaped, interleaved. Meaning they go in opposite directions on alternating laminations. Otoh, if you put all the "E" lams together, and then all the "I" lams together, the I would be like a cap on the end of the E... that flat spot is the gap. Usually it is adjusted for appropriate magnetic characteristics with a thin spacer. This is also how a "swinging choke" is constructed, fyi.

Actually, both a swinging choke and regular smoothing choke are constructed that way.  The difference is that the spacer is much thinner on a swinging choke, which is designed to deliver the maximum inductance at low DC current, at the expense of inductance at high DC current at full load, when less inductance is needed to maintain optimum inductance.  A smoothing choke has a wider gap to reduce saturation, but that limits the inductance at low values of DC.  So a smoothing choke has less inductance for the same coil and core, but there is less variation with the DC passing through it.

An a.c. transformer has the laminations interleaved,  both power transformers and audio transformers not designed to carry unbalanced DC through any of the windings.

I disassembled the driver transformer in a Ranger and interleaved the core, then used a blocking cap and plate resistor with the paralleled sections of the 12AU7.  Returned the plate resistor (about 10K or 30K, don't recall which) to the +600 volt supply instead of the +250 volt supply, to make up for voltage drop in the plate resistor.  That drastically improved the low frequency response of the Ranger.

I did the same with the mod transformer in an Eico 730 modulator that I use with my 10m rig.  Added blocking cap and 40-Henry 120MA modulation reactor (an ADC unit actually designed for the purpose).  With some modifications to the speech amp section, that modulator is actually flat down below 30~.  High frequency response is flat up to about 10k.

[WBear2GCR]

Don,

Fwiw, I have some chokes that are interleaved lams... dunno what exactly the idea is/was there...

With respect to the way you did the thing with the higher B+ and the dropping resistor, the beauty of the "parafeed" is that the choke on the plate takes the place of a resistor, but has a much lower DCR, while the AC Z is much higher, and increases as you go higher in frequency! So with the lower DCR you do not need to go to a higher B+, standard B+ should give you ample swing, and not reduce the quiesent current. The choke does not need to be very large in low power circuits, like a pair of 12AX7s in parallel...

Try it some time.

   

           _-_-WBear2GCR

[K9ACT]

>A single ended speech amp that feeds an interstage tranny through a 500mfd cap?
Seems a bit too large, unless the input Z to the tranny is very very low indeed.

It's 4 ohms.


>What exactly is this speech amp? Homebrew? A commercial amp?

It is a kit from canakit.com, a pair of TDA2003's in bridge config.


>It has DC at the output? What power is it? What is on the collector of the output transistor(s)?? etc...

12VDC on the output collectors.  20W audio into 4ohms

>The squeal is feedback or a parasitic.
Or RF getting into the silly EQ... probable.

Seems to be the latter.  An RF choke and bypass fixed it.

>Suggest you google for "parafeed" ... a good read.

Did that and found it very interesting.             

The good news is that the EQ seems to be working now.  Just need to confirm with someone that it actually helps and to tweek it in.

Not a good night for that.

Thanks,

Jack

[WBear2GCR]

Jack,

FB OM!

 

Now, you say you have those plasty-pak amps in bridge config?? Why not put the transformer in the center of the bridge, so you do not need a blocking cap? 

No DC that way... but I'd have expected that you would not have any DC at the output of the amp, unless that is a single-ended supplyl?? Does it want to run off a single-ended supply?? Why not go +/- and trim or servo out the DC component??
More better I would think...

                  _-_-WBear2GCR

[k4kyv]

With respect to the way you did the thing with the higher B+ and the dropping resistor, the beauty of the "parafeed" is that the choke on the plate takes the place of a resistor, but has a much lower DCR, while the AC Z is much higher, and increases as you go higher in frequency! So with the lower DCR you do not need to go to a higher B+, standard B+ should give you ample swing, and not reduce the quiesent current. The choke does not need to be very large in low power circuits, like a pair of 12AX7s in parallel...

But the choke takes up much more space than a resistor, and in the Ranger, space is a premium.  Since the +600 is available, and the total  current drain is no more than about 20 m.a., the  resistor is a better solution.  It just needs to be mounted so that it doesn't cook any nearby components with the heat it gives off.

Besides, for good low end response, you need loads of inductance for low-current applications. The rule of thumb used for modulation reactors is at  least 8 Henries of inductance for every 1000 ohms of load impedance.  Don't remember exactly what the rated load impedance for that pair of 12AU7 sections in parallel is, but I'd guess at least 7,500 ohms.  That would require a choke of at least 60 Hy @ 20 milliamps.  Not something you would easily find at the local electronics store, and it would probably be much larger than the Ranger driver xfmr.

[WBear2GCR]

Yeah Don, no doubt no room in the Ranger...

Btw, I looked at the pix of ur shack and xmiter... wow, very impressive set up.
Just wondering though, why no glass on the rack panels to see those toobes??

        _-_-bear

[k4kyv]

There is a perforated screen, which allows ventilation as well as a view of the finals.  Not sure if it shows up very well in the photo.  No fans anywhere in the rig.  Everything is 100% convection cooled.