# ESL bias voltage and sound output data

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

#### bentoronto

Here are sound output measures of a single Dayton-Wright cell with two levels of voltage, 1500 and 3000 v. Mic about an inch from diaphragm.

The output to the DW cell from a DW transformer appears as the yellow curve. This curve was taken (1) with no ESL load on the transformer and (2) may not correctly represent the top frequency drop-off since the load presented to the transformer by the REW test set-up does not present a very high input impedance.

Changing bias on this rather widely spaced, high diaphragm resistance cell from 1500 to 3000 requires about 20 minutes to reach close to maximum output.

Ummm, looks like ESL bias theory is roughly correct.

Ben

#### Attachments

• DW 1500 and 3000 v.jpg
56.7 KB · Views: 440

#### bentoronto

And with the same set-up using a Dayton-Wright transformer (with the output curve as shown in the previous post... this is with a quite flat curve for input to the transformer to roughly 15 kHz), again about 1 inch mic spacing, a single Dennesen tweeter cell (4 used in my curved tweeter array).

Curves are 750 v versus 1500 v bias. This cell gets to full bias charge in a few seconds.

Ben

#### Attachments

• Dennesen 750 and 1500 v.jpg
36 KB · Views: 432

#### invisible force

Hi Ben,
the high voltage resistor string Mike used failed with all of my four XG8MK3 - use a 20Meg 1Watt high voltage type from vishay instead..if you want to improve the design give each of the ten cells a 20 Meg resistor of its own (some work, I know)..problem with one big resistor for all cells is :let one cell leak slightly and the the bias for all cells will dropp..the >>5kV bias he claimed is a myth (with or without gas), after the resistor it will not be higher then ~6kV..5KV bias is on the safe side for no discharge noise (distortion is not the problem, coating is highly resistive and for my experience will last forever), with 5kV and 20 Meg charge up is fast <5Minutes..
Could you measure impedance curve of your transormer without load and with a 800pF capacitor? I can recommend the thinnest cloth, you can find for dust protection.. best regards, Philipp

#### bentoronto

Hi Ben,
the high voltage resistor string Mike used failed with all of my four XG8MK3 - use a 20Meg 1Watt high voltage type from vishay instead..if you want to improve the design give each of the ten cells a 20 Meg resistor of its own (some work, I know)..problem with one big resistor for all cells is :let one cell leak slightly and the the bias for all cells will dropp..the >>5kV bias he claimed is a myth (with or without gas), after the resistor it will not be higher then ~6kV..5KV bias is on the safe side for no discharge noise (distortion is not the problem, coating is highly resistive and for my experience will last forever), with 5kV and 20 Meg charge up is fast <5Minutes..
Could you measure impedance curve of your transormer without load and with a 800pF capacitor? I can recommend the thinnest cloth, you can find for dust protection.. best regards, Philipp
Much good advice. Actually, roughly what Mike Wright told me in 1975: use lotsa resistors everywhere. The XG-8 cells (at least prior to 1975) have two resistors hidden in the framework.

However, I seem to recall (from 1975) that the voltage at the cells with a few high-Ohm resistors in series (as best as I could measure it with a 100 meg probe) wasn't materially drooped from the bias power supply (at the time, an oil furnace ignition transformer.... I know how bad that is).

Great happiness: questions like the losses due to a grill cloth can be answered definitively today with REW and 2 minutes of time and a cheap mic.

I'll check the step-up transformer with an 800 pf cap, soon.

Ben

Last edited:

#### bentoronto

DATA POINT - resistance of a diaphragm

I have a 6400 v bias supply (when loaded by the following).

It is connected to 6 XG-8 cells in parallel but first, each cell is fed through a 50 megOhm resistor. (And, before the voltage gets to the diaphragm, another big resistor (inside) which i have no means of reaching.)

The power supply is 6400 v (EMCO black brick).

The diaphragm terminal on a few cells measured at the bias terminals, is about 4400 v.

So, following 50 megOhms, we have a 100 megOhm probe in parallel with a DW cell. The measured voltage is about the same as if the cell wasn't there (OK, actually even higher... just error of measurement); in other words, the cells didn't seem to be leaking any current, let alone 6 of them in parallel.

Hope that conclusion isn't faulty.

Ben

#### geraldfryjr

You might want to build jig like the one I have presented before
in order to get some very accurate measurements,

A TEST JIG FOR FINDING ESL STEP-UP TRANSFORMER PARAMETERS

It has been well worth the \$5 or \$10 investment for me!!!

You can build the divider with a higher resistance if you like, I just happened to only have enough resistors on hand to make this 20Meg version after building one for my variable HV supply.

I can measure the output of the step-up directly into the computer using REW or any other RTA, FR, or FFT program with extreme accuracy!!

And it also provides for highly accurate DC measurement's with your simple average DMM as well!

Here are some pictures of the two versions I have made and some sample measurements.
The values on the DMM are in KV.

Do, use an opamp buffer in order to isolate the meters impedance form the output of the voltage divider, the result will be the most accurate measurements you can get for AC and DC measurements.

These where actual measurements from the out put of an Antek AS-1206 as a step up transformer and the 8th photo shows what happens to the voltage coming fromit at the onset of core saturation.

The glitches you see in the REW graph's were caused from my onboard sound system and my add-on sound cards don't not do this.

Enjoy, FWIW !!

jer

#### Attachments

• THD at 20Vpeak-14.14Vrms 6V primary.jpg
276 KB · Views: 96
• AS 1206 Saturation at 20v-10v-5v peak input 6v and 12v primary.jpg
278.1 KB · Views: 89
• Onset of core satuation at 40Vrms 240Hz.jpg
134.4 KB · Views: 87
• AS 1206 output test 2400hz sine.jpg
357.7 KB · Views: 89
• IDILING AT 10 KV.jpg
108 KB · Views: 85
• IDLING AT 11KV.jpg
132.4 KB · Views: 100
• H V Boards.jpg
81.4 KB · Views: 98
• HIGH VOLTAGE DIVIDER.jpg
67.1 KB · Views: 354
• Test Jig In Action.jpg
143.7 KB · Views: 353
• Transformer Test Jig Board Layout.jpg
843.5 KB · Views: 374
Last edited:

#### bentoronto

Good test design for the audio interface. But the input side of the transformer shouldn't need high input impedance test instruments (because it is driven by a low output impedance amp).

Till now, I've measured output voltages with a 1(??) megOhm ACVTVM. One megOhm might load a 1:100 transformer, dunno.

To use REW, I'll need to revamp my 100,000,000:100,000 probe (which is a 1000:1 voltage divider) because the computer hardware will be inaccurate when in parallel with 100 kOhms.

More troublesome to me is the question of the location of grounds on the two sides of ESL matching transformers and the possibility of cooking gear when I fail to get the grounds figured out right.

High voltage work is kind of like quantum physics in that measuring stuff can be tricky.

Ben

Last edited:

#### geraldfryjr

The board I described and made has provisions for both!

It measures the output of the amp (input to the transformer) and the output of the transformer at the same time, as long as your software allows for recording of two channels at the same time.

Or you can use two meters to verify that your input level is exactly were it should be without having to stop and switch the clip leads for every single measurement.

My divider has a 10,000:1 division ratio in order not to exceed a the input voltage level to the sound card of 1V to 2V peak on AC voltages (and DC) as well for 10Kv and 20Kv respectively.

The bottom resistance of the divider is at 2K and has a trimmer for fine adjustment of initial resistor tolerance errors.

Even with just my DMM the error is not very significant at all with its 10Meg input impedance.

By this calculator the error comes out to only .03% with my DMM and even less with the common opamps that have a input resistance into the Gigaohm range,

Voltage Divider

A common multimeter can have an input resistance as low as 20K and this would cause an error as high as 10% lower than it really is.
As do cable capacitance's as well for AC measurements.

The very high input impedance of your average unity gain FET input opamp eliminates any of this error, and, any loading effects that changes with frequency that occur from the cables capacitance and the sound cards input impedence.

Any measuring device needs to have a higher impedance than the device it is measuring and the higher this is the more accurate the measurement is, for the most part.

Originally I deliberately designed it with a low output impedance of 2K thinking that this would be low enough to not worry about such errors.

But earlier test showed that I was still having a significant amount of error at the high end of the scale with a drooping curve when I knew that my amplifier produced a Ruler Flat FR.

And adding the buffer opamp's fixed this.

You can even power the opamp(s) from a 9V battery if you wish, the circuit is so simple and doesn't require but maybe 10ma to power the opamp(s).

I created this circuit especially for making these types of measurements.
This is a very small investment to have, to get some very accurate measurements easily.

Cheers!!

jer

P.S. The divider I used in my HV variable supply was set at 1000:1, and I have four other dividers on my test jig setup at different ratio's for various voltage ranges and power levels of various amplifiers that are being tested.
This very same test jig is used to impedance measurements of the transformers primary and speakers as well depending on how it is set up.

Last edited:

#### bolserst

Till now, I've measured output voltages with a 1(??) megOhm ACVTVM. One megOhm might load a 1:100 transformer, dunno.
It all depends on the transformer. Usually 1Mohm imparts minimal loading of the transformer, but transformers with a higher ratio of leakage inductance to winding capacitance sometimes show loading even with 10Mohm.

More troublesome to me is the question of the location of grounds on the two sides of ESL matching transformers and the possibility of cooking gear when I fail to get the grounds figured out right.
The best approach I have found is to use a battery powered differential probe to avoid the grounding issues. If using a single ended probe, connect the ground of your probe to the center tap of the transformer secondary, and measure each phase of the transformer output separately (ie front stator, rear stator) and then use REW to perform complex sum of the two measurements. With most transformers you will find the two phases have slightly different HF response due to different winding parasitic.

…Changing bias on this rather widely spaced, high diaphragm resistance cell from 1500 to 3000 requires about 20 minutes to reach close to maximum output.
Ummm, looks like ESL bias theory is roughly correct.
Just curious how you set/determined the bias voltages? Your plots only show a 2dB – 3dB difference in output when doubling the voltage should result in 6dB increase. Did you completely discharge supply and ESL before setting the 1500V level and measuring?
SPL-vs-Vbias

Last edited:

#### bentoronto

Thanks for guidance.

Jer - 10,000:1 probe sure sounds nice.

bolserst - Connecting the probe ground to centre tap does help albeit at twice the effort, if you want to eyeball both sides of the winding.

In as much as my interests were elsewhere last week, I was just in passing noting the crude relationship between crudely assessed bias voltage and crudely assessed sound output. Since I am surprised when I find any regularity in data, I thought I'd create a new thread just to document this.

I "measured" the bias voltage by adjusting the input voltages to my EMCO high voltage black-brick using a \$10 VOM (EMCO claims the output is linear with the input). Golly, measuring within 50% of theoretical expectations I consider (almost) a bulls-eye hit. Do you think a guy like me could ever pass an engineering course?

Ben

Last edited:

#### bolserst

Connecting the probe ground to centre tap does help albeit at twice the effort, if you want to eyeball both sides of the winding.
The purpose of connecting the probe ground to the center tap is to avoid introducing a loading on one half of the winding vs the other. It is the only way I found to get accurate HF measurements with a single ended probe that matched those from a friend's fancy differential setup. You're right though...it is added effort.

...measuring within 50% of theoretical expectations I consider (almost) a bulls-eye hit. Do you think a guy like me could ever pass an engineering course?
Hey wait a minute...with your references to use of Bell Labs equipment, I thought you were an engineer
Often just getting data that trends in the same direction as theory seems like a great triumph.
In the case of bias voltage, I have found ESL output to track theory surprisingly closely as shown in the the previous post I linked to.

Just for fun(since I stumbled on the data) I took the NF response sets I collected for the current source thread
http://www.diyaudio.com/forums/plan...rrent-vs-voltage-drive-esl-7.html#post2366501
and compared output vs bias voltage for frequency > 200Hz (ie above diaphragm modes)

Top plot shows response as measured with near field mic placement.
Bottom plot shows overlay if each curve is adjusted down by the theoretical increment for ratio of bias voltage relative to the lowest (870V) curve.
The results were pretty satisfying, with all curves stacking nicely on top of each other.

#### Attachments

• Vbias_vs_SPLinc.png
162.3 KB · Views: 225

#### bentoronto

Hey wait a minute...with your references to use of Bell Labs equipment, I thought you were an engineer .
off-topic

Nah, just a human-factors psychologist. But I was real friendly with the guy who ran the world's largest anechoic chamber (certainly at that time), Jim West, initial developer of electret technology. Funny, he figured electrets might be useful in microphones. Imagine that. Nice spot for me to fiddle with motional feedback.

And fooling with REW today (and thanks to bolserst for re-writing the manual) I noticed the names of two people who hung around the "Human Information Processing Department" when I was there: Tukey and Schroeder. Maybe more Bell names from that period in REW, but guess I am one of those folks who can't remember the 60's too well....

Insider secret: if you want to hear what Hal's original Bicycle Built for Two sounded like, you need to hear it on a KLH6 the standard for our department, which was created in our department a few years before I got there.

Ben

Last edited:

#### bentoronto

DATA POINTS

Bias voltage measured at cell connectors (one at each end of the cell). Loudness measured with mic a few inches from cell, 4500 Hz. dB figures nbot calibrated to anything. Noise about 25 dBA below test signal.

2070 v: 53.0 dB
3980 v: 56.7 dB

This is lower than figures provided by bolserst in previous post and below theory but more than you'd think from my first charts.

BTW, here's some further bias warm-up info.

I found my single cell in air asymptotes near full loudness at a bias voltage of 2070 v after maybe 5 minutes and fully charged to the last tenth of a dB or two at 12-15 minutes. After a few seconds of bias starting, down about 3 dB and within a dB of full activation after another minute.

Mike Wright says somewhere to give the speakers over-night to get charged.

Ben

Last edited:

#### anthonybisset

Not wanting to redirect the thread but maybe my ongoing troubleshooting will be useful to other Dayton Wright owners. I recently picked up upgraded XG8's with leaf tweeters and an XIM-10 interface module for a really good price ((the working condition was unknown)).

On getting them home I connected a crown xls1000 amp to the XIM-10 and attached one speaker. Turning on the XIM-10 immediately produced a very strong 50hz or 60hz hum {from inside XIM-10}. Strong enough that it made me wonder if there was a problem. After 20 minutes the single attached speaker was still incredibly quiet even driven with 175watts, while the humming was disturbing (and louder than the speaker!). After an hour or even 2 hours the speaker was putting out maybe 55dB (yes, still very quiet). Looking at schematics and inside at the PCB I noticed one of the large ~780mfd input caps had leaked but it wouldn't make any sense for this to be causing humming issue. I swapped channels anyway to see if there was a difference.. Nope. Reading in the XIM-10 manual there was a note about toggling the power switch a bunch of times to clean the relay if a humming sound was heard… I tried this for awhile and it did seem to help but only maybe 1 out of 50 toggles of the switch produced a quiet operating XIM-10 (no hum); but after maybe 60 seconds it would start humming loudly again. Eventually my friend suggested it might be dirty power relay contacts. I pulled the relay apart and cleaned both sides of the contacts with rubbing alcohol. Problem solved!

On opening the XIM10 to clean the dirty relay I found a crossover board, (that looked like an addition to the unit), had detached from the double-sided foam that was used to adhere it to the main PCB and was laying nervously against main board components. I carefully, without touching anything, re-attached the crossover board using zip ties to hold it safely in place.

Okay so far so good, but the problem of slow/questionable charging was still in play. Not knowing how long the XG8's should take to charge up and lore of them taking a day to be at full power I wasn't too worried about the very slow ramp up in volume that was occurring. However when the one plugged in speaker wasn't loud the next day (probably only 70dB), I started to wonder if there was a problem with the speaker or the XIM10. Without large resistors on hand I wasn't able to divide down the xim10 output for my multimeter, so I've been gathering data the old fashioned way.

Here is the data I've collected (so far),

1.) a single speaker plugged in takes about 3 or 4 days to reach a reasonable living room volume. (~80dB)

2.) having both speakers plugged in doesn't make much of a difference to charge up time… The speaker that was charged for 3 or 4 days seemed to remain charged even after a disconnect for an hour. (this suggests the bias circuit is not requiring both speakers be present to operate correctly as it works at the same speed with one or both speakers plugged in and each speaker's volume is only related to how long it's been plugged in). On adding the second speaker, it charged up even more slowly, i think it'll be at a reasonable level (~80dB) after 5 days given it's current progress.

3.) The bias control nob doesn't seem to make any difference to the performance of the system. I leave it set at 3/4 in case it is actually.

4.) The crown xls1000 amp produces 175watts into 8 ohms and 350 into 4ohms. Even with a +4dB input at full scale the output of the longest plugged in speaker (10 days now) is only maybe 83dB.

5.) The cells on the newest plugged in speaker seem to be charging at different rates… I think this happened with the first plugged in speaker as well. They all seem to be operating similarly given enough time to charge (days).

I'm hoping the problem is in the XIM10 vs the XG8's as opening them looks like a real hassle (the grill cloth is glued down underneath the strap) and pulling them open is possible but finding time is another thing.

I've seen talk in the thread about resistors strings or resistors going bad. I'm not clear on where these resistors are located or what kind are suggested for replacement… Would these be causing extremely slow charge time?

If XG8's can actually put out 100+dB (or provide headroom above that) I'd like to use them in studio and small room performative context as the midrange is very nice for mixing (skeptical of using them for mastering at this point, but maybe I'll find with mods to the XIM that they could get there)…

6.) For the first hour, when an uncharged speaker is plugged in I can put my ear next to the xim-10 and hear some higher frequency humming (not the 60hz relay sound)… It seems like it's working harder for a bit and then relaxes… Of course the output of the speaker is still whisper quiet after an hour

Any help resolving this super slow charge time would be appreciated.

I had a really nice experience mixing on a friends pair of XG10's in the past and hope to keep exposing people to the Dayton Wright ESL magic.

If the cells look mostly clear, is that an indication the SF6 is depleted? Does SF6 block light at all or have any opaqueness?

Last edited:

#### anthonybisset

Somehow I read the entire thread before posting and didn't catch that it is a general ESL discussion not focused on Dayton Wright... Oops.

#### invisible force

Well, the great Daytons...long,long story..I have no time now, but will go on the next days...for now :
- the high voltage socket give bad contact, simply drop a little 1 cm long wire (.5mm diameter when I recall right) in each contact before plugging together..
- the bias resistors in the speaker are prone to fail, get a 22-68Meg 1W resistor from vishay and replace it (cut back cloth, foil, you will see the string)
- charge up should be ~10 Minutes (little effect afterwards)
- bias supply also tends to fail, look around here for HF bias supply circuit, you need 5-6kV..
Philipp

#### Attachments

• circuit XG10.pdf
334 KB · Views: 80

#### bentoronto

I'm physically (and mentally) far remote from my DW system at the moment but thought I'd offer a word or two.

What a person in your circumstances is buying is certain major pieces (such as transformers and speakers) along with a bunch of small fixable headaches typical of any complex system. The Dayton-Wright major pieces are durable and the minor pieces (like the nutty interlock system likely imposed on the manufacturers... maybe or maybe not wisely) can be repaired, replaced, or tossed.

First step in a fault-tree search would be to determine if the speakers are intact and working OK. The rest should be pretty simple fixing.

Pity that like the phone wiring diagrams of old, the speaker schematics hide rather than reveal.

You have great sound ahead of you.

Ben

Ben

#### anthonybisset

Searching around about HV supplies for the Dayton Wright I came across this post by Paul Young regarding measurement.

---

"I have not been able to locate a circuit diagram for the EHT unit - but I do remember that it was a conventional diode / capacitor ladder type, composed of approx. 8 to 12 repeats (can't remember) of the same components. My memory of these parts was that they were fairly easy to purchase ceramic (radial lead) caps rated at 1500 volts or higher and that the diodes were low current plastic body units rated above 1500 volts also. The important construction detail was that the ladder was built on a fixture that suspended all of the components in air while they were soldered. No PCB was used to avoid HV arcing on its surface. All solder joints were kept smooth to avoid corona generating sharp points. The completed assembly was then potted in a plastic box by pouring a melted ”wax” (not really just wax - it was a product formulated for HV potting) into the box while the ladder of parts was suspended in the center - well away from the walls. The input voltage to the EHT unit was 90 VAC. Most IM10's also had a bias adjust power resistor in series with the primary side of the EHT transformer (either 0-2.5K ohms or 0-10K ohms - 25 Watt) to allow for reducing the 90 VAC to lower values - but factory practice and recommendation was to set this resistor at zero ohms and run the system at full voltage. This produced an operating bias voltage of approx. 10.2 kV (+/- 10%).

NOTE - this is very difficult to measure! The only tool that we found effective for these tests was an electrostatic voltmeter. You cannot use a HV Probe that performs 1000/1 division. All probes of this type that I am aware of will SIGNIFICANTLY drag down the actual voltage due to their low impedance. Electrostatic meters look like they came from Dr. Frankenstein's lab (huge wooden boxes, big swinging needles, etc..) but they do not load the circuit. A probe can pull the EHT down to half of its real output in some cases. "

-Paul Young

----

Seems the SF6 is not a gimmick.

#### invisible force

my experience :
5-6kV is reliable without discharge noise on the long run...but the Voltage has to reach the cells..again, if high voltage bias string in the speakers fail..no output..
the gas is not preventing discharge noise..but with gas and outer foils you lose 50% of inner details soundwise and get bad decay time in the bass and no highs..arcing is more provoqued by clipping amps, because air is more easy arcing with higher frequencies, then it will arc with or without gas...so : the gas is a gag..

#### bentoronto

my experience :
snip
..but with gas and outer foils you lose 50% of inner details soundwise and get bad decay time in the bass and no highs..arcing is more provoqued by clipping amps, because air is more easy arcing with higher frequencies, then it will arc with or without gas...so : the gas is a gag..
Since "inner details" has not yet been defined by any standards organization, further details on your post would be welcome.

There are posts which are "in my theoretical opinion",
"based on a lot of listening to this speaker...",
"a number of people have said to me...",
"I have seen discussions....", or some which are
"based on my measurements..."