4450人阅读
王小草深度学习笔记（13）
标签（空格分隔）： 王小草Tensorflow笔记
笔记整理者：王小草
笔记整理时间：日
代码原文请见github：
当使用Tensorflow训练大量深层的神经网络时，我们希望去跟踪神经网络的整个训练过程中的信息，比如迭代的过程中每一层参数是如何变化与分布的，比如每次循环参数更新后模型在测试集与训练集上的准确率是如何的，比如损失值的变化情况，等等。如果能在训练的过程中将一些信息加以记录并可视化得表现出来，是不是对我们探索模型有更深的帮助与理解呢？
Tensorflow官方推出了可视化工具Tensorboard，可以帮助我们实现以上功能，它可以将模型训练过程中的各种数据汇总起来存在自定义的路径与日志文件中，然后在指定的web端可视化地展现这些信息。
1. Tensorboard介绍
1.1 Tensorboard的数据形式
Tensorboard可以记录与展示以下数据形式：
（1）标量Scalars
（2）图片Images
（3）音频Audio
（4）计算图Graph
（5）数据分布Distribution
（6）直方图Histograms
（7）嵌入向量Embeddings
1.2 Tensorboard的可视化过程
（1）首先肯定是先建立一个graph,你想从这个graph中获取某些数据的信息
（2）确定要在graph中的哪些节点放置summary operations以记录信息
使用tf.summary.scalar记录标量
使用tf.summary.histogram记录数据的直方图
使用tf.summary.distribution记录数据的分布图
使用tf.summary.image记录图像数据
（3）operations并不会去真的执行计算，除非你告诉他们需要去run,或者它被其他的需要run的operation所依赖。而我们上一步创建的这些summary operations其实并不被其他节点依赖，因此，我们需要特地去运行所有的summary节点。但是呢，一份程序下来可能有超多这样的summary 节点，要手动一个一个去启动自然是及其繁琐的，因此我们可以使用tf.summary.merge_all去将所有summary节点合并成一个节点，只要运行这个节点，就能产生所有我们之前设置的summary data。
（4）使用tf.summary.FileWriter将运行后输出的数据都保存到本地磁盘中
（5）运行整个程序，并在命令行输入运行tensorboard的指令，之后打开web端可查看可视化的结果
2.Tensorboard使用案例
不出所料呢，我们还是使用最基础的识别手写字体的案例～
不过本案例也是先不去追求多美好的模型，只是建立一个简单的神经网络，让大家了解如何使用Tensorboard。
2.1 导入包，定义超参数，载入数据
（1）首先还是导入需要的包：
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import sys
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
（2）定义固定的超参数,方便待使用时直接传入。如果你问，这个超参数为啥要这样设定，如何选择最优的超参数？这个问题此处先不讨论，超参数的选择在机器学习建模中最常用的方法就是“交叉验证法”。而现在假设我们已经获得了最优的超参数，设置学利率为0.001，dropout的保留节点比例为0.9，最大循环次数为1000.
另外，还要设置两个路径，第一个是数据下载下来存放的地方，一个是summary输出保存的地方。
max_step = 1000
learning_rate = 0.001
dropout = 0.9
data_dir = ''
log_dir = ''
（3）接着加载数据,下载数据是直接调用了tensorflow提供的函数read_data_sets,输入两个参数，第一个是下载到数据存储的路径，第二个one_hot表示是否要将类别标签进行独热编码。它首先回去找制定目录下有没有这个数据文件，没有的话才去下载，有的话就直接读取。所以第一次执行这个命令，速度会比较慢。
mnist = input_data.read_data_sets(data_dir,one_hot=True)
2.2 创建特征与标签的占位符，保存输入的图片数据到summary
（1）创建tensorflow的默认会话：
sess = tf.InteractiveSession()
（2）创建输入数据的占位符，分别创建特征数据x，标签数据y_
在tf.placeholder()函数中传入了3个参数，第一个是定义数据类型为float32；第二个是数据的大小，特征数据是大小784的向量，标签数据是大小为10的向量，None表示不定死大小，到时候可以传入任何数量的样本；第3个参数是这个占位符的名称。
with tf.name_scope('input'):
x = tf.placeholder(tf.float32, [None, 784], name='x-input')
y_ = tf.placeholder(tf.float32, [None, 10], name='y-input')
（3）使用tf.summary.image保存图像信息
特征数据其实就是图像的像素数据拉升成一个1*784的向量，现在如果想在tensorboard上还原出输入的特征数据对应的图片，就需要将拉升的向量转变成28 * 28 * 1的原始像素了，于是可以用tf.reshape()直接重新调整特征数据的维度：
将输入的数据转换成[28 * 28 * 1]的shape，存储成另一个tensor，命名为image_shaped_input。
为了能使图片在tensorbord上展示出来，使用tf.summary.image将图片数据汇总给tensorbord。
tf.summary.image（）中传入的第一个参数是命名，第二个是图片数据，第三个是最多展示的张数，此处为10张
with tf.name_scope('input_reshape'):
image_shaped_input = tf.reshape(x, [-1, 28, 28, 1])
tf.summary.image('input', image_shaped_input, 10)
2.3 创建初始化参数的方法，与参数信息汇总到summary的方法
（1）在构建神经网络模型中，每一层中都需要去初始化参数w,b,为了使代码简介美观，最好将初始化参数的过程封装成方法function。
创建初始化权重w的方法，生成大小等于传入的shape参数，标准差为0.1,正态分布的随机数，并且将它转换成tensorflow中的variable返回。
def weight_variable(shape):
"""Create a weight variable with appropriate initialization."""
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial)
创建初始换偏执项b的方法，生成大小为传入参数shape的常数0.1，并将其转换成tensorflow的variable并返回
def bias_variable(shape):
"""Create a bias variable with appropriate initialization."""
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial)
（2）我们知道，在训练的过程在参数是不断地在改变和优化的，我们往往想知道每次迭代后参数都做了哪些变化，可以将参数的信息展现在tenorbord上，因此我们专门写一个方法来收录每次的参数信息。
def variable_summaries(var):
"""Attach a lot of summaries to a Tensor (for TensorBoard visualization)."""
with tf.name_scope('summaries'):
mean = tf.reduce_mean(var)
tf.summary.scalar('mean', mean)
with tf.name_scope('stddev'):
stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
tf.summary.scalar('stddev', stddev)
tf.summary.scalar('max', tf.reduce_max(var))
tf.summary.scalar('min', tf.reduce_min(var))
tf.summary.histogram('histogram', var)
2.4 构建神经网络层
（1）创建第一层隐藏层
创建一个构建隐藏层的方法,输入的参数有：
input_tensor：特征数据
input_dim：输入数据的维度大小
output_dim：输出数据的维度大小(=隐层神经元个数）
layer_name：命名空间
act=tf.nn.relu：激活函数（默认是relu)
def nn_layer(input_tensor, input_dim, output_dim, layer_name, act=tf.nn.relu):
"""Reusable code for making a simple neural net layer.
It does a matrix multiply, bias add, and then uses relu to nonlinearize.
It also sets up name scoping so that the resultant graph is easy to read,
and adds a number of summary ops.
with tf.name_scope(layer_name):
with tf.name_scope('weights'):
weights = weight_variable([input_dim, output_dim])
variable_summaries(weights)
with tf.name_scope('biases'):
biases = bias_variable([output_dim])
variable_summaries(biases)
with tf.name_scope('linear_compute'):
preactivate = tf.matmul(input_tensor, weights) + biases
tf.summary.histogram('linear', preactivate)
activations = act(preactivate, name='activation')
tf.summary.histogram('activations', activations)
return activations
调用隐层创建函数创建一个隐藏层：输入的维度是特征的维度784，神经元个数是500，也就是输出的维度。
hidden1 = nn_layer(x, 784, 500, 'layer1')
（2）创建一个dropout层，,随机关闭掉hidden1的一些神经元，并记录keep_prob
with tf.name_scope('dropout'):
keep_prob = tf.placeholder(tf.float32)
tf.summary.scalar('dropout_keep_probability', keep_prob)
dropped = tf.nn.dropout(hidden1, keep_prob)
（3）创建一个输出层，输入的维度是上一层的输出:500,输出的维度是分类的类别种类：10，激活函数设置为全等映射identity.（暂且先别使用softmax,会放在之后的损失函数中一起计算）
y = nn_layer(dropped, 500, 10, 'layer2', act=tf.identity)
2.5 创建损失函数
使用tf.nn.softmax_cross_entropy_with_logits来计算softmax并计算交叉熵损失,并且求均值作为最终的损失值。
with tf.name_scope('loss'):
diff = tf.nn.softmax_cross_entropy_with_logits(labels=y_, logits=y)
with tf.name_scope('total'):
cross_entropy = tf.reduce_mean(diff)
tf.summary.scalar('loss', cross_entropy)
2.6 训练，并计算准确率
（1）使用AdamOptimizer优化器训练模型，最小化交叉熵损失
with tf.name_scope('train'):
train_step = tf.train.AdamOptimizer(learning_rate).minimize(
cross_entropy)
（2）计算准确率,并用tf.summary.scalar记录准确率
with tf.name_scope('accuracy'):
with tf.name_scope('correct_prediction'):
# 分别将预测和真实的标签中取出最大值的索引，弱相同则返回1(true),不同则返回0(false)
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
with tf.name_scope('accuracy'):
# 求均值即为准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
tf.summary.scalar('accuracy', accuracy)
2.7 合并summary operation, 运行初始化变量
将所有的summaries合并，并且将它们写到之前定义的log_dir路径
# summaries合并
merged = tf.summary.merge_all()
# 写到指定的磁盘路径中
train_writer = tf.summary.FileWriter(log_dir + '/train', sess.graph)
test_writer = tf.summary.FileWriter(log_dir + '/test')
# 运行初始化所有变量
tf.global_variables_initializer().run()
2.8 准备训练与测试的两个数据，循环执行整个graph进行训练与评估
（1）现在我们要获取之后要喂人的数据.
如果是train==true，就从mnist.train中获取一个batch样本，并且设置dropout值；
如果是不是train==false,则获取minist.test的测试数据，并且设置keep_prob为1，即保留所有神经元开启。
def feed_dict(train):
"""Make a TensorFlow feed_dict: maps data onto Tensor placeholders."""
xs, ys = mnist.train.next_batch(100)
k = dropout
xs, ys = mnist.test.images, mnist.test.labels
return {x: xs, y_: ys, keep_prob: k}
（2）开始训练模型。
每隔10步，就进行一次merge, 并打印一次测试数据集的准确率，然后将测试数据集的各种summary信息写进日志中。
每隔100步，记录原信息
其他每一步时都记录下训练集的summary信息并写到日志中。
for i in range(max_steps):
if i % 10 == 0:
summary, acc = sess.run([merged, accuracy], feed_dict=feed_dict(False))
test_writer.add_summary(summary, i)
print('Accuracy at step %s: %s' % (i, acc))
if i % 100 == 99:
run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
run_metadata = tf.RunMetadata()
summary, _ = sess.run([merged, train_step],
feed_dict=feed_dict(True),
options=run_options,
run_metadata=run_metadata)
train_writer.add_run_metadata(run_metadata, 'step%03d' % i)
train_writer.add_summary(summary, i)
print('Adding run metadata for', i)
summary, _ = sess.run([merged, train_step], feed_dict=feed_dict(True))
train_writer.add_summary(summary, i)
train_writer.close()
test_writer.close()
2.9 执行程序，tensorboard生成可视化
（1）运行整个程序，在程序中定义的summary node就会将要记录的信息全部保存在指定的logdir路径中了，训练的记录会存一份文件，测试的记录会存一份文件。
（2）进入linux命令行，运行以下代码，等号后面加上summary日志保存的路径（在程序第一步中就事先自定义了）
执行命令之后会出现一条信息，上面有网址，将网址在浏览器中打开就可以看到我们定义的可视化信息了。：
Starting TensorBoard 41 on port 6006
(You can navigate to http://127.0.1.1:6006)
将在浏览器中打开，成功的话如下：
于是我们可以从这个web端看到所有程序中定义的可视化信息了。
2.10 Tensorboard Web端解释
看到最上面橙色一栏的菜单，分别有7个栏目，都一一对应着我们程序中定义信息的类型。
（1）SCALARS
展示的是标量的信息，我程序中用tf.summary.scalars()定义的信息都会在这个窗口。
回顾本文程序中定义的标量有：准确率accuracy,dropout的保留率，隐藏层中的参数信息，已经交叉熵损失。这些都在SCLARS窗口下显示出来了。
点开accuracy,红线表示test集的结果，蓝线表示train集的结果，可以看到随着循环次数的增加，两者的准确度也在通趋势增加，值得注意的是，在0到100次的循环中准确率快速激增，100次之后保持微弱地上升趋势，直达1000次时会到达0.967左右
点开dropout，红线表示的测试集上的保留率始终是1，蓝线始终是0.9
点开layer1，查看第一个隐藏层的参数信息。
以上，第一排是偏执项b的信息，随着迭代的加深，最大值越来越大，最小值越来越小，与此同时，也伴随着方差越来越大，这样的情况是我们愿意看到的，神经元之间的参数差异越来越大。因为理想的情况下每个神经元都应该去关注不同的特征，所以他们的参数也应有所不同。
第二排是权值w的信息，同理，最大值，最小值，标准差也都有与b相同的趋势，神经元之间的差异越来越明显。w的均值初始化的时候是0，随着迭代其绝对值也越来越大。
点开layer2
点开loss，可见损失的降低趋势。
（2）IMAGES
在程序中我们设置了一处保存了图像信息，就是在转变了输入特征的shape，然后记录到了image中，于是在tensorflow中就会还原出原始的图片了：
整个窗口总共展现了10张图片（根据代码中的参数10）
（3）AUDIO
这里展示的是声音的信息，但本案例中没有涉及到声音的。
（4）GRAPHS
这里展示的是整个训练过程的计算图graph,从中我们可以清洗地看到整个程序的逻辑与过程。
单击某个节点，可以查看属性，输入，输出等信息
单击节点上的“+”字样，可以看到该节点的内部信息。
另外还可以选择图像颜色的两者模型，基于结构的模式，相同的节点会有同样的颜色，基于预算硬件的，同一个硬件上的会有相同颜色。
（5）DISTRIBUTIONS
这里查看的是神经元输出的分布，有激活函数之前的分布，激活函数之后的分布等。
（6）HISTOGRAMS
也可以看以上数据的直方图
（7）EMBEDDINGS
展示的是嵌入向量的可视化效果，本案例中没有使用这个功能。之后其他案例中再详述。
本文主要使用手写数字识别的小案例来讲解了如何初步使用Tensorflow的可视化工具Tensorboard。
&&相关文章推荐
* 以上用户言论只代表其个人观点，不代表CSDN网站的观点或立场
访问：87910次
积分：1702
积分：1702
排名：千里之外
原创：78篇
评论：45条
(1)(3)(14)(3)(34)(2)(1)(7)(10)(1)(2)
(window.slotbydup = window.slotbydup || []).push({
id: '4740887',
container: s,
size: '250,250',
display: 'inlay-fix'TensorBoard 是 TensorFlow 官方推出的可视化工具，可将模型训练过程中的各种汇总数据（summaries）展示出来，包括：
标量（scalars）
tf.summary.scala(tensor.op.name, tensor)
图像（images）
音频（audio），视频（video）
计算图（Graphs）
数据分布（Distributions）、直方图（Histograms）
tf.summary.histogram(tensor.op.name, tensor)
嵌入向量（embeddings）
1. tensorflow 中的接口
tf.summary.scalar(‘mean’/’stddev’/’max’/’min’)
tf.summary.scalar(‘dropout_keep_prob’, keep_prob)
“标记”待可视化的变量
tf.summary.scalar('loss', loss)
tf.summary.histogram('w1', w1)
tf.summary.histogram('w2', w2)
定义 merge all 操作（参数为空，则表示对 default graph 中所有 summaries 进行 merge ）：
merged = tf.summary.merge_all()
实例化 FileWriter，将待可视化的变量序列化到本地；
writer = tf.summary.FileWriter(logdir='xxx',
sess.graph)
for t in range(N)
summary_str, ... = sess.run([merged, ...], feed_dict={x:..., y:...})
writer.add_summary(summary_str, t)
最后在命令行界面，切换到 logdir 指定的文件夹位置，使用如下的命令：
& tensorboard --logdir=xxx
仔细阅读命令行输出的日志信息，会给出相关的网址和端口信息：
(You can navigate to http://xx.xx.xx.xx:6006)
本文已收录于以下专栏：
相关文章推荐
1. 数据类型不带小数点的数默认为 int32，带小数点的数默认为 float32；
1. optimizer.minimize 与 global_stepoptimizer 的搭配使用
1. 一个实例
relu = tf.nn.relu(tf.matmul(x, W) + b)
C = [...][db, dW, dx] = tf.gradient(C, [b, w, x])
tf.clip_by_value() 截断，常和对数函数结合使用
正则作用的对象是目标函数，如图对均方误差使用 l2\ell_2 正则：loss = tf.reduce_mean(tf.square(y-y_) + tf.contrib.layers.l2_regu...
0. What are the differences between tf.initialize_all_variables() and tf.global_variables_initialize...
我乐于助人，我很高兴我十、我我我  我 很激动!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
官方文档的说明，Transitioning to TensorFlow 1.01. 简单列举如下
tf.VARIABLES => tf.GLOBAL_VARIABLES
tf.all_varia...
tensorflow 下的概率分布函数，一般用于对变量进行初始化，这里的变量显然是指神经网络的参数（连接层之间的权值矩阵和偏执向量）。
标准高斯：tf.random_normal([2, 3], st...
TensorFlow
tf 下以大写字母开头的含义为名词的一般表示一个类（class）
1. 优化器（optimizer）优化器的基类（Optimizer base clas...
他的最新文章
讲师：王哲涵
讲师：王渊命
您举报文章：
举报原因：
原文地址：
原因补充：
(最多只允许输入30个字)Larry Krantz Flute Pages: Flutes/Piccs Could Harm Hearing - S.A. Wicks
FLUTES OR PICCOLOS COULD HARM YOUR HEARING
How to Protect Your Hearing When Playing Your Musical Instrument
Steven A Wicks
INTRODUCTION.
Most of us naturally think of our flutes and piccolos as the instruments of beauty and artistic expression. And, because they allow us to make such beautiful music we might have difficulty thinking them capable of harming us in any way, perhaps not even our hearing. In fact, they are doing just that to many of us. The cause is the loudness of the sounds they produce.
What you are about to learn could be somewhat disconcerting. Since the damage is permanent, it might even take some reflection to understand fully the implications for your current goals and routines. In any case, with the evidence about to be presented, we should no longer ignore the fact that the loudness of our instruments makes them potentially damaging to our ears. Do not let the beauty of your instrument's tone mislead you or cause you to make choices you may regret later in life.
Please take a few minutes to learn about the causes of hearing damage in musicians and how you can protect yourself, your children or students. Here is what you will find:
THE NEED FOR CONCERN ABOUT HEARING LOSS
* How Hearing Losses Affect Musicians
* Musicians and Others with Damaged Hearing
* Why Our Instruments Have Become So Loud
CAUSES AND WARNING SIGNS OF HEARING LOSSES
* The Cause Of Hearing Loss In Musicians
* How Can I Tell If My Instrument is Harming My Hearing?
PLAYING IT SAFE
* How Loud Are Our Instruments?
* Maximum Sound Levels
* Recommendations For Common Playing Situations
HEARING PROTECTION
* Sorting Out Your Hearing Protection Needs
* How to Select Adequate Hearing Protection
REFERENCES AND BIBLIOGRAPHY WITH WWW LINKS
THE NEED FOR CONCERN ABOUT HEARING LOSS
---------------------------------------
HOW HEARING LOSSES AFFECT MUSICIANS.
We may assume as long as we can hear our instruments and other people's voices that our hearing is healthy enough. We should be aware that hearing damage does occur in many segments of instrumental music. Hearing damage in musicians is from loud sounds and can be a painless, progressive process, going undetected for many years. It would not be until many of us reach the point of hearing muffling of voices, begin asking for clarifications in conversations, or develop a ringing in our ears that we might discover the full degree of our loss.
Hearing loss from loud sounds affects judgement of loudness, pitch and tone, which, of course, are major ingredients of our artistic endeavors. Losses begin in the higher frequencies, so higher pitched sounds are perceived as less loud compared to lower pitched sounds. As it progresses it would become increasingly more difficult to produce any aspect of our instrument's tone, such as the upper partials, since we could not adequately hear them. We would lose our judgement of balance between the higher and lower frequencies making comparisons of relative loudness more difficult. It would affect our judgement of the relative loudness of our own instrument with others in ensembles. Then, our pitch discrimination, even in those with perfect pitch, would be degraded. Those with only moderate losses do seem to develop compensating techniques, but their innate abilities are measurably reduced compared to fully hearing-capable individuals. As if that is not enough, those with hearing losses tend to be more stressful and reclusive, too.
MUSICIANS AND OTHERS WITH DAMAGED HEARING.
Of the 250 million people in the U.S., a staggering 28 million are estimated to be suffering from hearing losses and at least 10 million of those are known to have suffered some losses from loud sounds. Some researchers believe these numbers are far too low, that perhaps most of us have lost some hearing from exposure to the loud sounds that fill our modern world. Already in a number of studies of musicians, from rock through classical, hearing damage has been revealed. In most studies of orchestras, just about HALF the musicians have hearing damage from loud sounds and may have been exposed to sound levels that exceed sound safety standards by many times, daily(1)(2). Additionally, the damage could exist in many thousands of semi-professional musicians, students, conductors, and band directors, who also are known to have high daily exposures.
Responsible adults, parents, and teachers need to take the lead in protecting those too young to understand the dangers of loud sounds. Because of the early onset of damage in a series of exposures, younger players may sustain losses before becoming mature or informed enough to take the proper precautions for themselves.
WHY OUR INSTRUMENTS HAVE BECOME SO LOUD.
As flutists and piccolo players we should be especially aware that our instruments easily produce sounds loud enough to cause hearing damage even when played alone. We may recognize that their loudness is on a par with other orchestral instruments, but we tend to ignore the fact that they are only six inches from our ears while they make those loud sounds. That closeness makes them quite a bit louder. Then, some of us play in ensembles or use PA systems that raise sound levels further, all of which adds to the burden our ears must sustain.
Lately, our modern world thrives more and more on loud sounds. Loudness has become almost a requirement for successful, or even acceptable, performances. Very likely, over the years instrumentalists have also preferred instruments that are louder, choosing ones that "speak" easier, or that respond to a lighter playing technique. In recent years, orchestras and artists have discovered audiences more often prefer louder, more brilliant or dynamic sounds. They have responded by playing louder pieces more often, and other pieces more loudly. Manufacturers have responded by providing louder instruments to meet growing demands. Then, musicians just naturally played them louder as time passed, in a sort of mezzo-forte-creep, up the dynamics scale. Adding to the increase in loudness is the fact that as general loudness of all sounds increases, so do hearing losses. Those with progressively increasing losses will naturally tend to play louder to compensate. Increasing general loudness can be a possible sign of increasing hearing loss. By middle age the average male in the US has lost 25 to 30 dB in the higher frequencies.
CAUSES AND WARNING SIGNS OF HEARING LOSSES
------------------------------------------
THE CAUSE OF HEARING LOSS IN MUSICIANS.
Loud sounds from any source, including music, can cause musicians to lose their hearing. Whether from music, noise, machinery, headset radios, gunshots, etc., the intensity of loud sounds causes a breakdown of the delicate mechanisms and tissues of the inner ear. The first exposures may produce only a temporary desensitization and allow a return to normal hearing after resting. With repeated exposures, however, at least a portion of the damage becomes permanent. Audiologists and hearing conservationists call such losses Noise Induced Hearing Loss (NIHL). Since there is no known cure we need to take a cautious approach to conserving our hearing. It cannot be reclaimed, later.
Those suffering from NIHL often think they are "getting used to hearing loud sounds," but they are simply well on their way to becoming "hard of hearing." They have a NIHL related dip in their hearing response between 3000 and 6000 Hz. In the initial progression of the loss, audiograms, used to diagnose hearing ability, begin to show a dip around 4000 Hz. With continued exposures the dip widens and deepens. It may take as little as one to ten years to show significantly. The rate of development depends on the loudness, regularity and length of the individual exposures, and our individual genetic sensitivity to such damage.
Since most of the damage from a series of exposures occurs in the earlier years, it is all the more important to detect and stop it, before it progresses too far. Putting off protective procedures or diagnoses of suspected damage, will only lead to more severe damage. Although the rate of progression is often related to the level of the sound, damage continues until hearing is destroyed, as long as the exposures continue. There is a tendency in those suffering from it to expose themselves to higher than they often turn up the volume on PA systems, hi-fi's and headphones, thereby accelerating the rate of destruction.
A primary warning sign of NIHL is a temporary loss of hearing sensitivity, called a Temporary Threshold Shift (TTS), according to the Hearing Conservation Program Training Manual of the Centers for Disease Control (CDC)(6). We can sense a TTS as a tired, ringing, buzzing or dull feeling in our ears following an exposure to loud sounds. Our ears need rest between exposures -- some say as much as 14 hours -- but even with rest, permanent damage can still occur. We need to quickly recognize and avoid ear-tiring situations to prevent permanent damage.
We should also realize that when both ears are exposed equally, we lose a great deal of our ability to judge how tired our ears have gotten. Exposing both ears causes us to lose our reference to the rested condition of our hearing. Therefore, developing a mental awareness of the loudness of sounds is what is required. Waiting for physical sensations in our ears to warn us may be too late.
You may be thinking that your instrument does not make your ears very tired. To see how rapidly they do tire, try this one-ear test. Wait until your ears are well rested, say, overnight. Insert an earplug (one with an NRR above 20 dB works best) in just your left ear. Play your instrument normally. The earplug will allow your left ear to act as a "rested reference" for your right ear. If your hearing is normal, your right ear will begin to feel tired within a surprisingly short time.
HOW CAN I TELL IF MY INSTRUMENT IS HARMING MY HEARING?
You need to have your hearing tested to know its condition. An audiogram is the standard diagnostic tool used by medical professionals and audiologists. We all should have audiograms taken regularly to detect any bad practices we may have developed in regard to loud sounds. They can also detect the effects of diseases that might be curable. However, be aware that waiting until damage shows up on your audiogram may be too late. At least a portion, if not all, of the measured loss could have become permanent. If you see the NIHL dip appear in your audiogram around 4000 Hz, you need to get a medical opinion. If it appears to be NIHL, you need to cut your exposures. A combination of prevention AND audiometric testing is still the best course of action.
Anyone experiencing dullness, ringing, muffling, beating, or other unusual effects in either ear very often, should get a medical diagnosis. They can be warning signs of advancing hearing loss, and may also be caused by a number of diseases. However, if you have any of these effects after a loud sound exposure you are unnecessarily endangering your hearing. Remember, NIHL is permanent. You cannot decide to do something to restore your hearing, later. Not even hearing aids can restore the function of destroyed tissues and nerve endings in your ears.
PLAYING IT SAFE
---------------
HOW LOUD ARE OUR INSTRUMENTS?
The following table lists the loudness of a number of common sounds, as well as, showing the ranges of flutes and piccolos. It shows the destructive potential of the sounds, as measured in A-weighted decibels, dB(A). The A-weighting in this case tells us how potentially damaging the various loud sounds are, not exactly how loud all of them are to our ears. Sounds heavy in low frequency components will tend to seem louder than indicated, but may not be as destructive.
As the table shows, flutes and piccolos at mezzo-forte rival the destructive loudness of a train whistle only 500 feet away, even though the train whistle seems louder to our ears. They can easily exceed the destructive level of the whistle, too. In fact, the top dynamic of a flute or piccolo is at least several times louder in terms of destructive potential and may even seem as loud in terms of actual loudness.
However, when we say a sound is several times louder than another we are discussing our perception of it. Doubling the perceived loudness of a sound generally requires a 6 to 10 dB increase, depending on whose ears we are testing. However, the middle and inner ears are stressed by the physical intensity of the sound, in proportion to the motion of the air molecules on our eardrums, not necessarily by our perception of it. Those stresses increase by a factor of ten for each 10 dB increase in the frequency ranges of flutes and piccolos(3)(4)(5).
Let's compare the top-end dynamic of a flute to the sounds we would hear far away from most other manmade sounds, like at the Home in the Country, as shown in the following table. The environment there is a quiet 30 dB(A). If we played our flute, some of us would perceive the top-end dynamic, at 105 dB(A), to be at least hundreds of times louder. No doubt, this is a lot, but the physical sound on our eardrums is actually 32 MILLION TIMES more intense. For a piccolo, the top end is 158 MILLION TIMES more intense. When we compare those figures to normal conversation at three feet, which is only 3200 times more, we begin to see how really intense some manmade sounds have become. Use the table to compare your flute or piccolo to some common sounds and to recommended maximum exposure durations listed in CAPS.
LOUDNESS OF FLUTES AND PICCOLOS COMPARED TO OTHER SOUNDS
[Table above is a .gif image - click on right mouse button to Save As...]
As can be seen in the table, we often are subjected to loud sounds when we are pursuing recreational or work activities in our modern world. Some of the biggest offenders, not shown, are headphone radios, thriller movies, home theater systems, concerts, sporting events, machinery, recreational vehicles, lawn care equipment, and snowblowers, to name just a few. Even sitting next to the window in a jetliner can give you more than the safe daily dose of loud sound, all in a single cross country flight, and without even playing a single note of music.
The Flute and The Guitar - A Contrast.
Contrast the destructive potential of a flute in the upper mezzo-forte range, 88 dB(A) average, with that of an acoustic guitar, which generates a mere 75 dB(A) at the same dynamic. Flutists are subjected to 13 dB greater potential. As a result, flutes place 20 times the destructive stress on our ears and require cutting the maximum exposure by 20 times. If an acoustic guitarist played loud enough to reduce her or his safe exposure time to 10 hours, a flutist would run out of safe playing time in as little as 30 minutes at the same dynamic. If this makes playing of our flutes and piccolos seem a bit more daring than we once thought - it should. The closeness to our ears makes our flutes and piccolos louder, but more importantly, in this case, they are also pitched in the range where the most damage is done to our ears the quickest.
MAXIMUM SOUND LEVELS.
There is no doubt that the subject of hearing conservation is complex when tackled all at once. There are technical terms and physiological concepts to learn. However, it is no different than other important health related issues in that respect.
In the following material many of the complexities have been reduced to guidelines. In reality, the human body seldom responds to any potentially harmful process in exactly the way stated by guidelines. Guidelines necessarily understate the physiological complexities involved. It will take an evolving awareness on the part of each individual to understand how sounds affect their hearing. In addition, sound, being invisible and complex, is difficult for the casual observer to adequately characterize. Some skill needs to be acquired to perceive situations where dangers exist and to apply appropriate protective measures.
Since everything about hearing loss is not fully understood, even by the experts, we need to stay abreast of new developments to stay safe. For example, over the past several years, recommended exposure times for loud sounds have been cut dramatically as the result of re-evaluations of existing programs and hearing loss data.
Hearing Conservation Concepts.
There are two basic methods for protecting oneself from loud sounds. For most musicians the final result will not be all that different regardless of the chosen method. The recommended, and by far the easiest choice, is simply to use hearing protection. The other method requires one to know the sound levels to which they are exposed and to learn how to manage exposures to them. As was seen in the previous section, without adequate measurements, called sound surveys, this can complicate matters considerably. In either case, most musicians, from intermediate level through professional, will likely find they must use hearing protection for at least a portion of their playing. Understanding both methods will make selecting hearing protectors much easier. Both have the same starting point so we will begin our discussion by defining safe sound levels.
Experts say that below about 75 dB(A), sounds of almost any duration are safe, for most people(8). This level is just about twice as loud as normal conversation from 3 feet away. We must realize this sound level is not very loud compared to sounds encountered in our modern world and certainly not loud enough to accommodate the playing of our flutes and piccolos much above piano or mezzo-piano levels. In order to have exposures to louder sounds we can either use hearing protection or we can consult published exposure tables. The tables are used to trade exposure time for increased loudness and could be used if the sound levels are known.
According to The National Institute of Occupational Safety and Health (NIOSH), hearing damage begins at 85 dB(A), after 8-hours of daily exposure. This assumes repeated exposures over extended periods, perhaps as long as 30 to 40 years. This corresponds to approximately the sound level generated at a player's ears by a flute at an average mid-mezzo-forte dynamic.
Since we obviously need to play at higher dynamic levels, too, exposure tables say for each just detectable increase of 3 dB in sound level over 85 dB(A), the allowable exposure time is cut in half. At a sound level of 88 dB(A), the maximum time is cut to 4 hours, for 91 dB(A) it is cut to 2 hours, and so on(7). Times derived from this guideline are all MAXIMUMS, where damage is KNOWN to begin. There are some drawbacks to using this scheme. It is based on data for exposures primarily to continuous sounds and many of the time-trading rules for the higher sound levels are based on theories or animal studies, not totally on experience with humans.
The exposure trading scheme, presented above, is one of the most conservative to be proposed, but still may result in hearing damage in some sensitive people. One reason for this is our music has an impulsive sound characteristic, not a continuous characteristic. It has sound peaks that considerably exceed the average due to its natural, expressive dynamics. Peaks may exceed the average sound level by 8 to 16 dB. It is not known from experience with humans if maximum exposure times derived from this scheme will truly protect against the impulsive nature of music, particularly over a person's entire lifetime. Projections say that roughly eight percent of those exposed in this manner for 40 years will suffer losses significant enough to show on audiometric charts. Also, flutists may have longer playing careers than the one on which this guideline is based, but better data is just not available. Of course, cautious persons will not want to push the limits, but before we ponder the implications of this too much, we need to be aware of a "catch-all" safety net inserted by NIOSH. This safety net will be our primary, recommended method.
Method 1 - The Use of Hearing Protection.
NIOSH recommends that anyone exposed to sound levels of 85 dB(A), or above, for any duration during a day, should wear hearing protection. The protectors should reduce the sound level at our eardrums to 82 dB(A), 8-hour TWA, or below. We will show how this works in some detail, in a moment, but be aware that nearly all flutists and piccolo players would be included by virtue of the loudness of their playing. Only beginners with less developed tones might escape it, even though they perhaps should not desire to do so. The rest of us would require hearing protectors simply because we would equal or exceed 85 dB(A) average sound level at our ears when we play at mezzo-forte, or above. Audiometric testing is also required.
In addition, the absolute maximum sound level of 115 dB(A) must be observed. Beginning between 115 dB(A) and 120 dB(A), instantaneous damage occurs in many people. Damage is no longer proportional to the duration of a sound, but depends only on the level. It is not likely that such levels could be generated by flutes and piccolos playing alone, but in ensembles or through the use of PA systems such levels could be reached, routinely. Orchestras playing loudly can exceed 120 dB(A), for example. Hearing protection will keep most loud sounds from reaching this level.
Let's review Method 1. Most of us, if not all, are advised by NIOSH to use hearing protectors and to have periodic audiometric exams. Beginners might not need protection, but should be advised of the need as soon as they progress enough to have well developed tones or begin to play in the higher dynamics. Then, we should not expose ourselves to greater than 115 dB(A) levels, on any occasion, for any duration, in order to avoid instantaneous damage. It is possible to manage exposure times using a sound meter, but the management of trades between time and sound level may prove to be more risky.
Genetic Variations Affecting Hearing Conservation.
Before moving into Method 2, we need to point out that each individual, not surprisingly, has a genetically determined sensitivity to hearing damage. This means it is not even possible to say with complete certainty that any particular sound level, 75 dB(A), 80 dB(A) or 85 dB(A), for example, is the safe exposure limit for all individuals. It is still a good idea to have periodic checkups to be certain that your hearing is being protected by any guideline. Individual susceptibilities to damage may differ by as much as 30 to 50 dB, which is a huge variation. Some of us may not suffer ill effects at all at fairly high levels, but the process of proving it may needlessly endanger a portion of one's hearing.
In light of all the limitations on exposure times from loud sounds, it is easy to get the impression that neither the playing of our instruments or their design should in any way be thought of as having been genetically ordained safe. Many of us are pushing hard on our physiological limits by playing them without adequate hearing protection. Since most hearing programs are designed to get someone through a 30 to 40 year career, those who started young and intend to play a flute or piccolo, perhaps for 60 years or more, needs to be particularly conservative.
Method 2 - Exposure Limiting.
This method will allow trading exposure time for higher sound levels. The basic concepts were covered in the previous sections, and involve reducing exposure times above 85 dB(A) by half for each 3 dB increase in level. The information in this section is primarily for Method 2, but will be useful as background safety information for those planning to use Method 1. There will undoubtedly be times when those using Method 1 will be caught without their earplugs.
The next chart shows the maximum recommended daily sound exposures based on concepts developed by NIOSH(7). It uses 75 dB(A) as a safe level for any duration and 85 dB(A) as the maximum for 8 hours daily. The maximum durations at various sound levels are in the first two columns. The music dynamics have been added to illustrate APPROXIMATE playing levels. They CANNOT accurately substitute for sound measurements taken near your own ears, so you should only use them to help see the need for sound measurements or for exercising caution. This simply means if you are at all close to any limit you should either back off, make sound meter measurements or use hearing protection. The Hearing Protector columns were also added to help in choosing an adequate protector when open ear exposures are too short.
All stated exposure times are maximums for full 24 hour days, to be followed by 14 hours of rest. Sound exposures at various levels or for various activities must be totaled as percentages of the maximums to determine your Total Dose (covered in Figuring Your Daily Sound Dose, below). Sound levels should be measured at the ear using a calibrated sound meter or dosimeter (see Sound Meters, below).
[Table above is a .gif image - click on right mouse button to Save As...]
Dynamic indications are for illustrative purposes only and will not necessarily predict your own sound levels accurately. They are for solo players. Sound levels in ensembles could be significantly higher at each dynamic level.
** Music average or peak levels exceeding 115 dB(A) may cause permanent instantaneous damage regardless of exposure time. Some ensembles and PA systems can generate these levels on music peaks.
Figuring Your Daily Sound Dose.
All individual, daily, loud sound exposures must be totaled to give your Total Dose, including, practices, performances and recreational activities. A Total Dose of 100 percent is the maximum for each 24-hour day. To get the Total Dose for a set of exposures, figure the percentages for individual exposures, based on their maximum exposure times, and sum the percentages. Remember to use percentages of each maximum time in figuring your Total Dose, for example: 15 minutes of a 30 minute maximum is 50 percent, not 15 minutes. Those desiring more safety will not push toward the maximum times or a Total Dose of 100 percent.
* An Open Ear (without earplugs) exposure at 91 dB(A) for 2 hours (100 percent dose) uses up your entire daily allowance for all loud sounds above 75 dB(A). Earplugs that would reduce all other playing to below 75 dB(A) would be required for all other loud sound exposures. Use the NRR columns to find suitable earplugs based on the level and duration of the remaining sounds.
* An Open Ear exposure at 97 dB(A) for 15 minutes (50 percent dose), followed by 4 hours at 85 dB(A) (50 percent dose) will use up your entire daily allowance for all sounds above 75 dB(A). Earplugs that would reduce all other playing to below 75 dB(A) would be required for all other loud sound exposures. Use the NRR columns to find suitable earplugs based on the level and duration of the remaining sounds.
* An Open Ear exposure at 91 dB(A) for 1.5 hours (75 percent dose) is desired. However, you need to play at lower levels for longer than the remaining 25 percent of Total Dose will allow. You could use earplugs to reduce the exposure at the higher levels and have sufficient Total Dose left for the other playing, with Open Ears (see How To Select Adequate Hearing Protection).
RECOMMENDATIONS FOR COMMON PLAYING SITUATIONS.
In all of the following discussions, the conclusion likely will be that hearing protection makes the most sense. There are many variations of playing style possible, but all are dominated by the same overriding factor: the short maximum exposure times at the higher sound levels. In the forte range, for example, Total Dose gets used very rapidly, as the chart shows. If you find your other exposures too limited by this, you should be using hearing protection, either all the time or at least when practical, to conserve Total Dose. If you experience any of the hearing effects in Warning Signs, below, you need to cut exposures more. Everyone should see Room Acoustics, Warning Signs and Piccolo Players.
Beginners and Casual Players.
Beginners and more casual players who do not play above mezzo-forte have up to two hours of open ear playing time each day (without earplugs). This might be extended if a sound meter could be used to verify an average level of 85 dB(A), or lower. Those who play at higher dynamics should be using hearing protection.
Intermediate through Professional Players.
More advanced or intense players who spend very much time at higher dynamic levels will find they will use their Total Dose quickly, leaving too little safe exposure time for the rest of their playing. For example, either 15 minutes at forte (97 dB(A)) or 1.5 minutes at double-forte (106 dB(A)), would use 50 percent of your Total Dose (for the entire day). In effect, either exposure would cut all other times in The Maximum Sound Exposure Chart in half. It is unlikely that the remaining 50 percent would allow enough time to finish your daily playing. If it is not enough time, earplugs will be required, at least for a portion of your playing. See the next section on using a sound meter to verify your levels at the higher dynamics.
Sound Meters.
A sound meter would be required to measure your sound level, which becomes more important at the higher dynamics where the maximum times are measured in minutes. Small changes in loudness can have a large effect on the exposure time and Total Dose. However, unless your routines are fairly stable, a sound meter is not very handy to use. It can help you learn how loudly you are playing specific or representative pieces, drills, etudes, and scales, in some venues. In other venues, acoustical variations make measurements too tedious. The sound meter will not be very handy in many actual performance situations, which would have to be estimated from experience.
Using A Sound Meter.
Select a sound meter that will average sound levels over a period of at least several minutes. Measure average sound levels, not peaks or maximums. Mount your sound meter on a camera tripod quite near your right ear. Set it for A-weighting and take averages over the length of entire pieces, preferably over the entire playing period. Look up the sound levels in the chart and figure your dose in percent using the time over which you averaged the sound. If this is not equal to your total playing period, adjust the dose accordingly. Be sure your meter is calibrated and add the meter tolerance to the measured values to err on the side of safety.
Playing More Softly.
Playing more softly may be an option for some who exceed the exposure limits. If you are a solo player or can convince your ensemble to cooperate, this is a good first step toward safer playing. The 6 dB NRR column in The Maximum Sound Exposure Chart shows how the exposure times are increased for playing at half loudness. Half loudness will extend your maximum time by a factor of about four.
Amplification and Playing Softly.
If you use amplification, you can play more softly and turn up the PA system to compensate. To be effective, ALWAYS avoid sitting in front of your PA loudspeakers and keep monitor speaker levels low, too. Use a sound meter to determine if you are operating safely. Room acoustics will likely require you to take new measurements in each venue. If you must sit in front of the loudspeakers, use hearing protection.
Larger Ensembles.
Ensemble players are often at risk. Orchestra members typically have the greatest hearing losses of acoustic musicians. The high combined sound levels in medium to large ensembles may be in the mid-90 dB range, or higher, even at mezzo-forte. Even in orchestra pits small ensembles can reach these levels, too. Due to the inherent, expressive dynamics of music, peak sound levels can exceed 115 dB(A), the absolute maximum sound level. Instantaneous hearing damage begins to occur above that level. Your group's management should take responsibility for sound surveys, which most likely will point to the need for hearing protection. The sound surveys will be useful in selecting earplugs. If possible, have audio professionals take readings at the ears of most of the players in the various sections, especially those in or near the brass and percussion sections. Be aware that the plastic, sound-deflecting screens used in many orchestras provide varying amounts of protection from minimal to significant, and should not be rearranged without confirming their continued effectiveness by sound measurements. Be sure to include your practices, rehearsals, and recreational exposures in figuring your Total Dose.
Smaller Ensembles.
If you play in a smaller ensemble, and you do not sit in front of or too near a loud instrument, you need at least to consider the contribution of the room acoustics. If the sound level is significantly higher in ensemble than when you play alone you need to use a sound meter to determine how safe you are. If the sound level is only perceptibly louder in ensemble, cut the exposure times by two to four in the Open Ear or NRR columns. This will likely cut the times too much to be practical and require the use of earplugs. If your group or band has brass or percussion instruments, you should be concerned with the possibility of music peaks exceeding 115 dB(A), the absolute maximum sound level. Instantaneous hearing damage begins to occur above that level. Groups heavy in brass and percussion instruments will have significantly higher average sound levels, too.
Music Teachers, Conductors and Band Directors.
Music teachers, conductors and band directors are also likely to exceed their daily maximum sound exposures due to the combined sounds of multiple instruments. There exists a multitude of possible variations in room acoustics and ensemble configurations, making it very difficult to predict the combined sound levels at your ears without using a sound meter. Estimates place the levels about one entire dynamic level higher for the combined ensemble those shown in the chart for single instruments. Room acoustics could raise it further, producing pockets of higher sound levels.
Teachers who sit in medium to small practice rooms with only one or two students should cut the maximum exposure times in at least half and limit the time spent at higher dynamics. If you spend an entire day in such quarters it would be hard to not exceed the limits. Be cautious if you are in a small, live room. With multiple instruments or small rooms sound measurements are recommended.
Room Acoustics.
Room acoustics have varying effects on the sound that reaches our ears. Practicing in a dead room (with carpets, drapes or stuffed furniture to absorb the sound) will often cause us to play louder. This raises the sound level at the ears of flute and piccolo players, since most of the sound comes directly from the instrument in most surroundings. For solo players, cut the maximum exposure times at least in half for dead rooms. If you play in a live room, unless it is very small, use the chart values (see advice for music teachers, above). The small live room raises the sound level at your ears even though you may be playing slightly softer.
Piccolo Players.
Piccolo players are perhaps in more danger than flutists and most other players. Their top dynamic is very loud and they are pitched where the most damage is done the quickest. Those who sit next to them are also subjected to very high sound levels. Sound levels three feet away may range from 5 dB lower to equal to those measured at the player's ears, depending on acoustics. A piccolo is capable of producing such loud sounds at double-forte that the player could exceed maximum exposures within just 1 to 7.5 minutes. Such short exposures rapidly use up your daily Total Dose, even on shorter passages. You are advised to use hearing protection.
Warning Signs.
Musicians with ringing, buzzing, muffling, harshness, beating or any other odd sensations in either ear should see a medical practitioner or licensed audiologist, soon. Some players with sensitive ears may even experience warning signs at most sound levels above even 75 dB(A). They definitely should be using earplugs. Have audiograms made and look for NIHL patterns. Keep copies of audiograms for future comparisons. If you have tinnitus (continuous ringing) or any other chronic condition, in either ear, and it worsens with an exposure, you are exceeding your limits.
HEARING PROTECTION
------------------
HOW TO SELECT ADEQUATE HEARING PROTECTION.
Earplugs will very likely be the correct choice for you and especially if you cannot perform adequate sound surveys in all venues. They are recommended by NIOSH for sound levels typically produced by most players of flutes and piccolos, as pointed out in Method 1, above. Unless you are a beginner or casual player who does not often play above mid-mezzo-forte, or are one of a very few amplified players, you will need them, even for use in limiting exposures in Method 2, above. Remember, a good hearing conservation program will protect your hearing for your lifetime, not merely expend a portion of it over a period of your career.
Basic Types of Hearing Protectors.
There are two basic types of earplugs, noise suppression and "hi-fi."
As the name implies, the hi-fi type has a flatter response, with less roll-off (decrease) in the higher frequencies. Noise suppression earplugs can have a greater Noise Reduction Rating (NRR), but the roll-off is usually greater. There are earmuff protectors, although not hi-fi, that could be used in the studio and are handy when they must be removed and replaced, often, as in teaching. Wearing any device that does not have a stated Noise Reduction Rating (NRR), such as, any materials found around the home is not recommended for any loud sound exposure, including: audio headphones, cotton, paper or cigarette filters. They often do not have the required high frequency reduction. Be safe.
Choosing a Noise Reduction Rating (NRR).
The Maximum Sound Exposure Chart shows that for most of us, the controlling factor in choosing a Noise Reduction Rating (NRR) is the time spent in the higher loudness levels. For most of us, playing alone, the exposure times afforded by 12 dB NRR earplugs will be sufficient in the mezzo-forte to mid-forte range (8 to 24 hours), having extended our playing time by up to 16 times. For those who spend more time in the higher dynamics or in ensembles, higher NRR's may be required. The goal is to reduce the level at our eardrums to 82 dB(A), or below, for the equivalent of 8 hours per day. The four NRR columns in the center of the Maximum Exposure Chart are based on meeting that requirement.
Derating NRR's.
Earplugs receive a Noise Reduction Rating based on testing in a laboratory. It should tell us how well a hearing protector will perform when we use it. Protectors often do not perform as well in real life situations as they did in the laboratory and the stated NRR on the package may not be what your will experience. The correctness of fit and the amount of bone conducted sound through the face and head varies between individuals, causing some to experience less attenuation of the sound than others. NIOSH recommends derating the manufacturer's NRR's as follows:
* Earmuffs: reduce to 75 percent
* Formable earplugs: reduce to 50 percent
* All others: reduce to 25 percent.
For example, a formable, foam earplug with a rating of 30 dB would be derated to 15 dB. You may get better performance than is indicated here, if you take time to read the use instructions carefully and to apply your protector conscientiously.
You can use The Maximum Sound Exposure Chart to see how your proposed earplugs will extend your playing time. The central columns show earplug NRR's of 6, 12, 18, and 24 dB. If the NRR of your earplugs, after derating, is not shown, use the next lower NRR in the chart. For example, for 25 dB NRR, formable earplugs, use the 12 dB column. If you have any doubt about the fit of the earplugs, or their effectiveness, due to wear-and-tear, you should replace them as soon as possible.
Is a Higher NRR Always Better?
Most of us will not prefer to use earplugs that exceed our basic requirements by very much. As we lower the proportion of the sound coming through our outer ear, bone conducted sound in our face and head becomes a major contributor to the sound we perceive. For some players, bone conduction emphasizes the lower frequencies, even with good quality, hi-fi earplugs. Using earplugs that are adequate for the required daily exposures will minimize this effect. Of course, when doing activities like mowing the lawn, and for all other non-critical listening activities, use the maximum NRR.
Choosing a Type of Earplug.
If you want the best quality earplugs, with a minimum of distortion and a broad range of NRR's, you will need to visit an audiologist. If you need high NRR's, as for piccolos, you will probably need to do so also, as the noise reduction earplugs (with the highest NRR's, as shown below) will not be nearly as satisfying. If you are looking for good quality earplugs that may suffice at a lower cost, try the hi-fi, standard fit types. If you need the highest NRR's in a low cost earplug, the noise suppression types are the only choice, at present. Audiologists should know of the latest types available in the widest range of NRR's.
Getting Used to Wearing Your Earplugs.
Once you have your earplugs, allow a couple of weeks, or more, to adjust to them. Plan to make the transition when your performance schedule best allows it. You will have to relearn how your instrument sounds at the new, safer levels. Periodically check your tone with and without the earplugs and make the necessary adjustments. If you are not satisfied with the resultant effects on your tone, move up to higher quality earplugs or work with an audiologist to get what you need. Before giving up on a particular earplug, have someone check your tone at a distance, say equal to the distance to your average audience, under similar acoustic conditions, if possible. Some sounds that are prominent at the player's ears do not project very far. This is especially true when using hearing protectors.
Typical Earplugs.
The following earplug listing is a representative list of types. There are many other brands available, but this list should allow comparing various types for beginning your search. You can enter the keyword, "earplug," in an Internet search facility to see more choices. The author does not endorse any particular brand for any purpose and the data below may change without notice. If you have questions about which earplugs are appropriate, you should consult a licensed audiologist or medical professional.
Noise Reduction Types, with high NRR.
E-A-R, Model PL-102, Cabot Safety Corp., formable, soft foam earplugs. About US$2 per three pair or US$5 per 12 pair. Intended for noise reduction. NRR on package is 29 dB. Flatness: about 20 dB roll-off at higher frequencies.
Flents, Model #150, Flents Products, Inc., formable, soft earplugs. About US$2 per package. Intended for noise reduction. NRR on the package is 29 dB. About 20 dB roll-off at higher frequencies.
Flents, Model #195, Flents Products, Inc., formable, soft earplugs. About US$3 per 5 pair. Intended for noise reduction. NRR on the package is 22 dB. Estimated 20 dB roll-off at higher frequencies.
MABCO, Quiet Space, Model 21612, formable, soft earplugs. About US$4.25 per 6 pair. Intended for noise reduction. NRR on the package is 31 dB.
Silaflex 2, Model 903, formable earcaps. About US$3 per 3 pair. Intended for noise reduction. NRR on the package is 21 dB.
Hi-Fi Types, with moderate NRR.
E-A-R, Model ER-20, Cabot Safety Corp., Hi-Fi, soft plastic triple tips. About US$20 per pair. Intended for musicians. NRR on the package is 12 dB. Flatness: stated 15 dB at low frequencies, with approx. 10 dB roll-off at higher frequencies.
Etymotic, Model ER-15, (data not available at the time of writing).
Hi-Fi Types, higher quality, with various NRR's.
Consult an audiologist for options. Prices are US$100, and up.
Noise control types are available at home improvement centers, hardware stores (medium NRR's) and gunshops (higher NRR's). They range in price from about US$10 to US$30. NRR's range from about 15 dB to 29 dB.
______________
REFERENCES AND BIBLIOGRAPHY WITH WWW LINKS
------------------------------------------
References.
1 Teie, Paul U., M.M., M.S., F.A.A.A., Noise-induced hearing loss and symphony orchestra musicians: risk factors, effects, and management, Maryland Medical Journal, 1998, pp 13-18.
2 Daum, Miriam C., P.T., M.P.H., Hearing Loss in Musicians, Center for Safety in the Arts, New York, NY, 1988. See WWW site: :70/0/Artswire/csa/arthazards/performing/musnoise
3 Everest, F. Alton, The Master Handbook of Acoustics, 3rd Edition, TAB Books, New York, 1994, Audibility of Loudness Changes, p. 49.
4 Davis, Gary, and Ralph Jones, Sound Reinforcement Handbook, Second Edition, Hal Leonard Corp., Milwaukee, 1990.
5 Pierce, John R., The Science of Musical Sound, Revised Edition, Loudness, pp. 123-7, W. H. Freeman and Company, New York, 1992.
6 Centers for Disease Control, Hearing Conservation Program, Appendix A, Noise - Training Information, Noise, Sections A and D. See WWW site: http://www.cdc.gov/od/ohs/manual/hearing.htm#Sec 3
7 Criteria for a Recommended Standard, Occupational Noise Exposure, Revised Criteria 1996, DHHS (NIOSH) Publication No. 96璛XX, National Institute of Occupational Safety and Health, 1996. See WWW site: http://www.cdc.gov/niosh/noisecd.html
8 National Institutes of Health, Consensus Development Conference Statement, 76. Noise and Hearing Loss, January 22-24, 1990. See WWW site: http://text.nlm.nih.gov/nih/cdc/www/76txt.html
Bibliography and WWW Links.
Environmental Protection Agency (EPA), Part 211, Product Noise Labeling. See WWW site: http://www.epa.gov/docs/epacfr40//subch-G/40P0211.pdf
Hamernik, Roger P. and William A. Ahroon, , Interrupted noise exposures: Threshold shift dynamics and permanent effects, Plattsburgh State University of New York, 1997. See WWW site: http://137.142.42.15/ARL/Recent/Hamernik_1997b/Hamernik_1997b.html
The Health & Safety Executive, UK, The Noise at Work Regulations, 1989. See WWW site: http://www.measure.demon.co.uk/docs/The_Noise_at_Work_Regulations_1989.html
H.E.A.R. (Hearing Education and Awareness for Rockers), HEARNET web site: /text/mainframe.html
Hearing Health Institute, Music/Audio, 1997. See WWW site:
http://www.hei.org/hei/music.htm
National Hearing Conservation Association, Guidelines for Audiometric Baseline Revision Recommended by the National Hearing Conservation Association, February 24, 1996. See WWW site: /~nhca/pos1.html
National Hearing Conservation Association, NHCA Greets New Director of OSHA: Suggests Strengthening OSHA Noise Policy, March 2, 1994. See WWW site: /~nhca/pos7.html
National Hearing Conservation Association, Position Statement on National Institute for Occupational Safety and Health, 1995. See WWW site: /~nhca/pos2.html
National Hearing Conservation Association, Position Statement on OSHA, and the OSHA Noise Standard and Hearing Conservation Amendment 1910.95, October 12,1995. See WWW site: /~nhca/pos3.html
National Hearing Conservation Association, Position Statement on the Quiet Communities Act of 1997, March 19, 1997. See WWW site: /~nhca/pos4.html
National Hearing Conservation Association, Recommendations of the NHCA Task Force on Hearing Protector Effectiveness, March, 1995. See WWW site: /~nhca/pos6.html
National Institute of Occupational Safety and Health, Occupational Noise and Hearing Conservation, Selected Issues. See WWW site: http://www.cdc.gov/niosh/noise2a.html#PAR1
Occupational Health and Safety Administration, Standards - 29 CFR, 1910.95, Occupational Noise Exposure. See WWW site: http://www.osha-lc.gov/OshStd_toc/OSHA_Std_toc_1910_SUBPART_G.html
Olsen, Harry F., Music, Physics, and Engineering, Second Edition, Dover Publications, Inc., New York, 1967.
Oshex Associates, Inc., The OSHA Noise Standard, 1996. See WWW site: /oshanois.htm
SAFTEK Information Services, Noise in the Music Entertainment Industry, Basic Concepts, 1998. See WWW site: http://www.saftek.net/worksafe/noise_e2.htm
University of Washington, Protect your ears from loud noise to hold on to your hearing, . See WWW site: http://www.hslib.washington.edu/your_health/hbeat/hb950418.html
_____________________________
Copyright 1998, by Steven A Wicks. This document contains copyrighted information. No further publication, in whole or part, without the written permission of the author: Steven A Wicks, 3901 Zenako Street, San Diego, CA, USA, . Version 980526.
BE HEARING YOU.}